Computer Programable Hearing Aid

INTRODUCTION

The purpose of this manuscript is to provide the reader with a knowledge of which computer programs are available to the clinician for facilitating the selection and verification of hearing aids. The objective is to provide a general conceptual understanding of the theoretical basis behind each program, what features the program has to offer, and to share an appreciation for the "look and feel" of the programs. Since microcomputers can be used in so many areas related to hearing aids, perhaps it would be useful to describe what will not be covered. Several topics are beyond the scope of this text, such as: a review of software designed by hearing instrument manufacturers to fit their own programmable hearing aids, software that is related to instrumentation involving probe microphone devices, or software for business management. The Hearing Instrument Manufacturers Software Association's (HIMSA) product, NOAH, will not be reviewed, but interested readers can find out more about NOAH and related applications by seeing Robertson (1996). NOAH is software designed to integrate hardware and software modules covering a wide range of applications from a variety of manufactures. One prominent application involves programmable hearing aids which use a microcomputer to control a hardware interface device for setting the parameters of the instrument. This is certainly an important use for microcomputers; however, most hearing aids sold are not programmable. Here, a more generic approach will be taken. The emphasis is on selecting appropriate electroacoustic characteristics for hearing aids whether they are linear, nonlinear, programmable or not. Also, this paper is intended to be non-judgmental and informative, not a critical review or evaluation of program features. Indeed, one program to be described was developed by the author who can hardly be considered impartial.

A traditional literature search (i.e., Medline) was not particularly useful in identifying software associated with hearing aid fitting. Even when fairly recent articles were found, often it was doubtful whether the program was still commercially available for the platform for which it was designed (e.g., Apple IIe). Three or four years is a long time in the computer business. Identifying what software is "out there" and currently available, supported, and would run on a major platform (Macintosh or IBM PC compatibles, Mac or PC) was the goal, but it was also a hit or miss proposition. So, apologies are offered in advance for the author's ignorance in inadvertantly omitting someone's favorite program or creation. Also, since this is not intended to be a treatise on the scientific aspects of hearing aid fitting, material will be presented in a less formal tone.

The main portion of the text will begin with a general review which, hopefully, will point out the current need for including computer-based schemes into the selection and verification process, and which issues need to be addressed in hearing aid selection. The underlying purpose of much of the software is to take the theory from basic research and to apply those concepts to the fitting of hearing aids. For the reader who would like more details than this general overview provides, please see Studebaker and Hochberg (1980; 1993) and Valente (1994). A brief summary will be included of what computer hardware would be needed to support the software. It is assumed that the reader is somewhat familiar with basic computer terminology and concepts. Developing a working understanding of microcomputers is not easy. Attempts have been made to keep descriptions simple for the beginner yet complete and accurate for the reader with a more advanced understanding. A full explanation of all computer related terms, suitable for the novice, would be very lengthy and beyond the scope of the paper. Following the hardware section will be a review of the features offered by the various programs (see Table 1).

Table 1.

Hearing aid selection and verification software. Minimum system requirements are included

Program	Author(s) or Contact	System
MSUv3.1	Robyn Cox Department of Audiology and Speech Pathology Memphis State University 807 Jefferson Avenue Memphis, TN 38105 Phone:901-678-5831 E-mail: ausp.memphis.edu/harl	PC XT, 256 kB RAM, dual floppy
DSL3.1	Richard Seewald University of Western Ontario Communication Disorders, Elborn London, ON NBG I-HI Canada Phone: 519-661-3901	PC XT, 640 kB RAM. 1 floppy, Mac version of DSL 3.1
DSL 4.0 or DSL [i/o]	E-mail: ac.owu.urchh.oidua@lsd	Windows, 386 processor
IHAFF vl.la	IHAFF working group Dennis Van Vliet 17021 Yorba Linda Blvd, Suite 130 Yorba Linda, CA 92686 Phone: 714-579-0717:800-795-4288	PC XT, 450 kB RAM. 0.5 MB hard drive
Fig6 3.0 Rev L	Mead Killion or Toni Gitles Etymotic Research 61 Martin Lane Elk Grove Village, IL 60007 Phone: 847-228-0006	PC AT, 525 kB RAM. I floppy
HAS 2.5 SII	Robert de Jonge Department of Speech Pathology and Audiology Central Missouri State University Warrensburg, MO 64093 contact: Support Syndicate for Audiology 108 S. 12th Street Pittsburgh, PA 15203 Phone: 800-869-0758	Macintosh, HyperCard 2.2 or 2.3
SII	ANSI Draft Standard (ANSI S3.79, proposed) Chaslav Pavlovic Resound Corporation 220 Saginaw Drive Redwood City. CA 94063 Phone: 415-780-7867	PCXT, 128 kB RAM. 1 floppy
HASP 2.07	Harvey Dillon National Acoustic Laboratories 126 Greville Street Chatswood 2067, NSW, Australia Fax: 612-411-8273	PC AT. 384 kB RAM. 0.7 MB hard drive

HARDWARE AND SOFTWARE REQUIREMENTS

Most of the programs (also referred to as "applications") run under MS-DOS (Microsoft Corporation's Disk Operating System), the main operating system for the PCs. An exception is the most recent version of the Desired Sensation Level (DSL 4.0), also called DSL [i/o], where "i/o" is an acronym for "input/output." DSL [i/o] is a Windows program. A Mac version of DSL 3.1 is also available. None of the other applications were Windows based, but all ran without trouble when they were invoked from the File Manager within Windows 3.1. The DOS programs are all modest in terms of the power of the microprocessor required to run them, PC XT or AT with 640 kB (kilobytes) of random access memory (RAM) or less. If an older computer is available, it could be dedicated to running these programs. In most cases the programs could run from a single floppy, but all could be installed on a hard drive and used less than 1 MB (megabyte) of disk space. There was no difficulty installing them in their own directory or running them on a PC with an Intel 486 processor. Only the DSL 3.1 software modified the autoexec.bat file, and this just added a line referencing a directory for storing patient files. So, removing the software, if necessary, is not complicated, and you needn't worry about the installation interfering with the operation of your existing software. The autoexec.bat file contains a series of commands which are executed when the computer is first starting up ("booting"). Changing some of these commands, or the order in which the commands execute, can potentially interfere with the proper operation of programs. Therefore, it's best that the autoexec.bat file be left alone unless there is a good reason to change it.

Even though these programs were designed to run on a PC, there are two ways that they can also run on a Mac. The first method is by software emulation, and the second is by installing additional hardware in the Mac. This second approach essentially puts a PC on a card into one of the Mac expansion slots. All of the PC programs, except DSL 4.0, ran on a Mac 6100/60 PowerPC computer running SoftWindows 1.02a. This is a program that emulates the Intel 286 processor (PC AT class). It permits the Mac to run DOS 6.21, Windows 3.1, DOS-based programs, and Windows programs in standard mode. Speed is comparable to entry level 486 machines. SoftWindows 3.0 emulates the 486 processor and runs Windows 3.1 in enhanced mode and also runs Windows 95. No additional hardware is required. Since DSL 4.0 requires, at minimum, a 386 processor, this program needs to run in enhanced mode. Enhanced mode mainly refers to the ability of 386 class machines (including 486 and Pentium) to use hard disk space as RAM. This "virtual memory" allows larger programs to operate. It also allows more programs to run concurrently. The main disadvantage to using SoftWindows software emulation is speed. Programs run slower. The DOS Compatibility Card, installed in the Mac 6100/60 PowerPC, is a 66 MHz 486 DX2 computer with 8 MB of RAM memory. Its performance was the same as a similarly configured stand-alone PC. The DOS card is sold by Apple Computer, but other third-party manufacturers offer similar products.

All the figures of the DOS-based programs were taken from the Mac screen. The appearance of the programs was the same whether they were viewed on the Mac or PC screen. Adobe Photo-Shop 2.5 was used to convert the original displays (which were often in color) into black and white images that were appropriate for viewing within this text. This sacrifice lessens the esthetic appeal of the original programs, but at least gives the reader a feel for the features offered by the programs.

For the Macintosh software, the Hearing Aid Selection (HAS) stack (program) required a Mac and the HyperCard 2.2 application. HyperCard is a programming environment for the object-oriented HyperTalk programming language. The HyperCard program can be used to create stacks, much like a word processing program creates data files. The term "stack" is a metaphor for a stack of "cards." Each card can be thought of as a screen image (Figures 19 to ​ 33). HyperCard stacks take advantage of the Mac's intuitive user interface of buttons, pull-down menus, etc. Stacks can be run from HyperCard or the HyperCard Player application that is distributed free of charge with each Mac. HyperCard Player allows existing stacks to be run, but new ones cannot be created. Stacks can also be compiled into free-standing applications. The HAS and Speech Intelligibility Index (SII) stacks will run on an older Mac Plus, but they perform significantly better if run on a Mac with a faster processor (like a 16 to 25 MHz Motorola 68030 processor, or better). Currently all Macs being sold perform at these levels or better.

Audiogram data entry card for the SII calculations.

Options for parallel vents.

The Need For Computer-Based Applications

In the early 1970s, when the author was first exposed to hearing aid fitting, the selection and verification process was much more of an art than a science. This was necessarily so, since so little was known, or, at least, little was common knowledge in the clinical, audiological community. The entire concept of hearing aid prescription was viewed as questionable. It was doubted whether a hearing aid could be prescribed with the same degree of confidence as, for example, eyeglasses could be prescribed. If a prescriptive procedure was used, it was mainly the half-gain rule (Lybarger, 1944). More commonly, hearing aids were selected at random, usually three or four aids, from a pool of instruments available in the clinic. The patient wore stock rather than custom earmolds. Hearing aids were compared using a variety of speech perception tasks, or some variation of the Carhart procedure (Carhart, 1946). It was recommended that the patient purchase (from a different person, the hearing aid dealer) the hearing aid giving the best performance. When hearing aids were being compared, it was uncertain whether the listening environment, speech presented in a sound-proof booth, was a valid simulation of the real-world listening experience. Objective measurements of hearing aid performance were limited to 2-cc coupler measures. The differences between decibels of coupler gain, audiogram thresholds obtained in supra-aural earphones calibrated on a 6-cc coupler, and real-ear gain were not appreciated. The pronounced effects of earmold acoustics on the frequency response of the hearing aid were typically not assessed or controlled. No attempt was usually made to determine how the hearing aid was shaping the spectrum (i.e., frequency response) for the individual. The selection process was not logically connected to a prescription. Verification of the selection was a self fulfilling prophecy since the Carhart procedure was used both to select and verify. It was doubtful whether speech-based measures could reliably choose the best hearing aid (Shore et al, 1960; Thornton and Raffin, 1978). And perhaps the best hearing aid was not chosen if it was not included in the original group of hearing aids randomly selected for evaluation. Follow-up evaluation was problematic since the person fitting the hearing aid was not the one making the recommendation, and there was no guarantee that the recommendation would be followed or, indeed, that the audiologist's recommendation was worthy of being followed.

Increased Hearing Aid Complexity

In subsequent years, hearing aids became more complex. The effect of this complexity on the perception of sound became appreciated as a significant issue. Behind-the-ear (BTE) hearing aids were being replaced by in-the-ear (ITE) hearing aids. The 2-cc coupler could now be either an HA-1 or HA-2 coupler depending upon whether an ITE or BTE aid was being measured, respectively. The two couplers are slightly different physically and consequently measure different sound pressure levels (SPLs). In-the-canal (ITC) aids grew in popularity as did completely-in-the-canal (CIC) aids. Changing the location of the microphone, as worn on the listener, would alter the input to the hearing aid, and consequently the output. These real-ear effects would not be measured on the 2-cc coupler. Hearing aids situated deeply in the ear canal would produce an output SPL at the eardrum much greater than hearing aids with a more shallow placement. Coupler gain would be different from real-ear gain for the average ear. Individual differences in ear canal geometry and middle ear impedance meant that the relationship between 2-cc coupler gain and real-ear gain would be unique for every individual (de Jonge, 1996). Most certainly, hearing aids would perform differently on children versus adults, normal middle ears vs. pathological ears, and effects of normal variability within the population could be expected.

Electronically, hearing aids were becoming more complex. Originally, most hearing aids had linear amplifiers. The frequency response of the hearing aid was constant as the input level to the hearing aid changed, at least up to the point where saturation occurred. Two response curves would describe the aid: one below saturation where the hearing aid is in the linear range (the frequency response), and one at, and above, saturation (the saturation sound pressure level with a 90 dB input, OSPL90). Which frequency response was appropriate for a patient was a question that had a single answer. This is not the case for hearing aids with frequency-dependent compression. Depending upon values selected for compression ratio(s) and kneepoint(s), the frequency response of compression hearing aids can change dramatically with changes in the input level. Choosing a frequency response became a more complicated question. What response is needed at which input level? In addition the dispenser has been given more control over the hearing aid. In the past hearing aids may have had a low-frequency cut or an output limiting control. Now, the audiologist can select a hearing aid from dozens of possible matrices with a variety of trim pots. And, especially with programmable hearing aids, the audiologist has control over a bewildering array of parameters, for multiple channels and memories, that can create a multitude of different frequency responses. The audiologist faces the dilemma of being able to deliver almost any frequency response, but being uncertain of which will maximally benefit the patient. Consequently, knowing what to deliver (the hearing aid prescription) becomes a very important consideration.

Advances in Basic and Applied Research

In the last twenty years, much progress has been made in a number of areas. Hearing aid prescription techniques have been developed. Probably the most well known, widely used, and experimentally verified is that developed at the National Acoustic Laboratories in Australia (NAL-R, Byrne and Dillon, 1986; Byrne et al, 1990). Important acoustical effects have been defined: the role of the head, neck, torso, pinna, concha, ear canal and middle ear in transforming sound from the freeor diffuse field to either the entrance to a hearing aid microphone or the eardrum (Shaw, 1974; Shaw, 1980; Bentler and Pavlovic, 1989; Table 2). Differences between hearing aid output in the real ear versus the 2-cc coupler (real-ear coupler difference, RECD) have been measured, and it is possible to convert one to the other (Killion and Revit, 1993; Revit, 1994). The factors producing individual differences in real-ear gain and output are better understood. And, most importantly, equipment for measuring the in-situ performance of a hearing aid is common and readily available commercially. Real-ear probe-microphone systems can sample the SPL for a wide range of frequencies at the eardrum of individuals whether they are unaided (REUR, the real-ear unaided response) or wearing the hearing aid (REAR, the real-ear aided response). The real-ear insertion response (REIR) can be measured as the difference between aided and unaided listening (REAR - REUR). The real-ear saturation response (RESR) can be compared to the OSPL90 of the hearing aid in the 2-cc coupler.

Table 2.

A summary of acoustical corrections often made by hearing aid selection software hoacoustical measures.

Variable	Name	Comment
Microphone location	micLoc	Increased SPL at mic location due to baffle of effects of head and external ear; varies for BTE, ITE, ITC, CIC aids
Real-ear SPL	RESPL	SPL developed in a real ear canal, close to the eardrum
Real-ear aided response	REAR	The RESPL created during aided listening
Real-ear saturation response	RESR	The maximum RESPL, measured with the aid in saturation, 90 dB SPL RESR = OSPL90 + RECD − 3mm bore
2-cc coupler SPL	2ccSPLl 2ccSPL2 2ccSPL	SPL developed in the HA-1, 2-cc coupler, SPL developed in the HA-2, 2-cc coupler Refers to either 2ccSPLI or 2ccSPL2
18 mm long, 3 mm diameter earmold bore	3mm bore	Produces greater high frequency output in HA-2 coupler versus HA-1: 3mm bore = 0 dB for HA-1 coupler measures 2ccSPL2 = 2ccSPLl + 3mm bore
6-cc coupler SPL	6ccSPL	SPL developed in the NBS-9A coupler used to calibrate circumaural earphones
2-cc coupler frequency response	2ccCR	Gain as a function of frequency for an HA-1 or HA-2 coupler
Output saturation sound pressure level	OSPL90	The maximum 2ccSPL, measured with the aid in saturation, 90 dB SPL OSPL90 = RESR + RECD − 3 mm bore
Real-ear coupler difference for an HA-1 coupler	RECD	RECD = RESPL − 2ccSPL Load impedance of ear canal and middle ear increases output, especially for higher frequencies in the real ear
Real-ear dial difference	REDD	REDD = DialReading - RESPL The difference between the audiometer dial reading (DialReading, in dB HL) and the RESPL
Real-ear unaided response	REUR	The difference in SPL developed by the reference mic and the RESPL due to external ear resonances
Real-ear insertion response	REIR	Real-ear insertion gain frequency response REIR = REAR - REUR
NBS-9A. 6-cc coupler	6cc	Converts 6-cc coupler SPL to RESPL RESPL = 6ccSPL + 6cc
Audiometric zero, circumaural earphones	ANSI	Converts dB HL to dB SPL in the 6-cc coupler, according to ANSI S3.6-1989 6ccSPL = dB HL + ANSI
ER-3A insert earphone	ER3A	Converts dB HL to dB SPL in the HA-1, 2-cc coupler, according to ANSIS3.6-1989 2ccSPL = dBHL + ANSI
Coupler response for flat insertion gain	CORFIG	Converts REIR to 2ccCR CORFIG = REUR - RECD + 3 mm bore - micLoc 2ccCR = REIR + CORFIG
The opposite of CORFIG	GIFROC	Converts 2ccCR to REIR GIFROC = −CORFOG REIR = 2ccCR + GIFROC
Audiogram threshold	dBHL	Conventional audiometric threshold measured in dB HL
Most comfortable level	MCL	MCL is usually viewed as the target output for amplified average conversational level speech, measured in dB HL, RESPL, etc.; MCL is often calculated from dBHL, or could be measured directly.
Loudness discomfort level	LDL	LDL is usually set to slightly less than RESR or OSPL90; LDL is often calculated from dBHL, or could be measured directly.

The characteristics of the average conversational speech spectrum have been measured for different vocal efforts (Pearsons et al, 1976; Pavlovic, 1993). The relationship between the speech spectrum and normal audibility is known so that the dB SPL of speech in the sound field can be compared to an audiogram measured in hearing level (dB HL). Techniques like the articulation index (AI) and speech intelligibility index (SII) can be used to make useful predictions. Without actually performing speech audiometry it is possible to predict the effects of hearing loss (i.e., different audiometric configurations) on speech intelligbility. It is possible to anticipate how modifying the speech spectrum through changing the frequency response of the hearing aid will alter speech recognition. It is also possible to predict how amplified environmental noise might interfere with speech perception (ANSI S3.79, proposed).

So, information is available that would make it possible for the audiologist to take a much more structured approach to the hearing aid selection and verification process. For example, after the audiogram is obtained, a set of calculations specified by a prescriptive procedure would generate a target REIR. The effect of earmold modifications and venting could be addressed (Dillon, 1985; Dillon et al, 1992). The effect of the REIR on speech intelligibility could be calculated. If a satisfactory result is obtained, then correction factors could be applied to each frequency to convert the REIR to a target 2-cc coupler frequency response. The correction factors would adjust for the effect of hearing aid microphone placement, the average REUR and RECD, and type of coupler (HA-1 vs HA-2). Experimental data could be used to generate a target RESR based upon average loudness discomfort levels (LDLs) for individuals with varying amounts of sensorineural hearing loss (Kamm et al, 1978; Pascoe, 1988). The target RESR, again based upon the average ear, could be converted to a OSPL90 curve for a 2-cc coupler. Once the frequency response and SSPL have been identified, an appropriate BTE aid could be ordered, or a matrix could be selected for ordering a custom ITE/ITC/CIC hearing aid.

For a more individualized fit, measurements of the patient's own loudness perceptions (i.e., LDL or most comfortable level, MCL) could be made. The patient's own REUR and RECD could be used to develop customized corrections. When the hearing aid arrives, any adjustments could be made (based upon 2-cc coupler measurements) to fine-tune the instrument in the absence of the patient. When the patient is wearing the aid, probe microphone measures would verify (or refute) that the real-ear fit is as expected based upon the initial prescription, and more fine-tuning could be performed. The effect of any residual error, of the measured REIR not matching the target, upon speech intelligibility could be calculated and a decision made whether to proceed with the fitting. A self report questionnaire could be used before and after a period of hearing aid use to evaluate subjective benefit from the hearing aid (Cox and Alexander, 1995; Fabry and Schum, 1994).

All of the above may seem complicated, too complicated to be worth the effort even if a better fit and, presumably, a more satisfied hearing aid user was the result. Or, at the very least, such attention to detail would be tediously time consuming. It would take away from time spent with the patient, and it would be impractical in the real world of hearing aid fitting. And this would be true if it were not for a remarkable invention that emerged in the late 1970s and has now become commonplace, the microcomputer. A microcomputer, running the appropriate software, can take the results of basic research and apply these results to the practical task of hearing aid fitting. The computer can reduce the complexity of adding correction factors, gain conversion, calculating speech intelligibility, etc. to a simple matter of entering some information about the patient and choosing the appropriate menu item. While the microcomputer cannot have empathy nor replace clinical judgment, perhaps it can be used as an assisting tool, allowing more science to be added to the process, freeing the clinician to pursue the art of hearing aid fitting.

SOFTWARE PACKAGES

The following is a brief description of the software packages available for assisting the clinician in the selection and verification of the electroacoustic characteristics of hearing aids. The description includes hardware requirements, theoretical basis of the procedure, and details of how the clinician interacts with the software. Generally, earlier procedures that were primarily designed to fit linear hearing aids are described first. More recent procedures tend to follow. Often, the more recent programs address issues involving frequency-dependent compression (including programmable) aids, issues such as target input-output functions and level dependent frequency responses. For this reason, version 4.0 of the DSL procedure is considered separately from version 3.1.

Memphis State University Hearing Aid Prescription (MSUv3.1)

Version 3.1 of the University of Memphis (formerly called Memphis State University) software comes on a single floppy disk which includes both the program and documentation. Hardware requirements are modest: PC XT class machine with 256 kB of memory, a hard drive or dual floppy, color or monochrome monitor, and a printer for text output. The documentation file can be printed to serve as a user manual. This manual is restricted to an explanation of how to use the program; there is no information about the (considerable) theory nor empirical research behind the MSU procedure. However, references are given to several articles which thoroughly explain the rationale and verification of the approach. To effectively use and understand the software, I found the articles to be very helpful and essential (Cox, 1983; Cox, 1984; Cox, 1985; Cox and Alexander, 1990).

The program is text-based, no graphics. Features are easily selected via the keyboard using pull-down menus. The user can create, save, and load patient files containing basic information such as name, date, date of birth, clinician, ear tested, comments, etc. All information relating to the hearing aid prescription and calculations made by the program can be saved. The information entered, a table containing the prescription, along with a worksheet for comparing three hearing aids, can be printed.

The principle used for selecting insertion gain is based on the long term listening range (LTLR). The LTLR range is defined by the patient's threshold at the low end and the HCL (highest comfortable listening level, formerly referred to as ULCL, the upper limit to the comfortable listening range) at the high end. Cox and Bisset (1982) found that the preferred listening level is close to the midpoint of the LTLR. If the LTLR is 30 dB or greater, conversational speech, produced with a raised vocal effort (70 dB SPL overall level), is amplified so that 1/3 octave band SPLs are positioned at the midpoint of the LTLR. For patients with a reduced LTLR, speech is amplified to a level 15 dB below the HCL. Insertion gain can be converted to coupler gain by applying the appropriate corrections for either a BTE or ITE aid.

The OSPL90 curve is selected with consideration given to both the loudness and clarity of speech:

OSPL90 = (HCL − 12dB) + speechPeaks

where the difference between the dB level exceeded by 1% of the speech peaks and the rms level is indicated by speechPeaks (Cox, 1988). The formula is designed to calculate an output level that would not exceed the user's comfort range, yet only allow 1% of the speech peaks to be distorted by clipping (harmonic distortion), if that is the method used for output limiting by the hearing aid. OSPL90 can be displayed for either an HA-1 or HA-2 coupler.

Thresholds can be entered for supra-aural or insert earphones (such as the Etymotic ER-3A). For the supra-aural earphone, thresholds can be expressed in either dB HL or the equivalent real-ear SPL (RESPL), called the sound pressure hearing level (SPHL). Insert earphone thresholds require SPHL. The HCL can be measured directly or it can be estimated. The method for measuring HCL uses a descending procedure and stimuli commonly available on standard audiometers. The patient is presented with either warble tones or 1/3 octave bands of pulsed noise. Loudness of these stimuli is comparable to the loudness associated with bands of speech. Initially, the presentation level is relatively high, close to LDL. The stimuli are reduced in level until the patient responds, indicating that the comfort range has been reached. A series of descents are performed, and the highest level producing a comfort response is HCL. The day-to-day reliability of this procedure is such that standard deviations (SDs) of test-retest differences are 5 to 6 dB. Long term (more than 17 months) SDs of test-retest differences for this procedure are greater, about 8 dB (Cox, 1989).

Since loudness testing can be time consuming and difficult, if not impossible, for some patients to perform, HCL can be estimated using one of two predictive techniques. The first uses regression equations to predict HCL from SPHL and take the following form:

where m ranges from 0.25 to 0.45 and b from 68 to 85 dB, depending upon frequency (Cox, 1988). Using this procedure produces errors in predicting the true HCL. Standard errors of estimate are about 9 to 10 dB. The second procedure was developed to reduce this error and requires that HCLs be measured at only two frequencies, 500 and 4000 Hz. The difference between predicted and measured HCL (called the residual) is calculated by the program and used to adjust HCLs at other frequencies. Standard errors of estimate using this technique were about 4 to 6 dB, which is comparable to the day-to-day test-retest reliability (Cox, 1989). So, if the patient can perform the HCL measurement, and if the dispenser wishes to take the time, a more accurate prescription can be made. Also, without making any loudness measurements, you can change HCL at 500 and 4000 Hz, based upon your judgment, to modify output levels for an individual you suspect to be able to tolerate more, or less, loudness. Figures 1 and ​ 2 show screen displays created by the MSUv3.1 program. For this example, HCLs and unaided sound field thresholds (USFTs) were predicted from threshold (THR). In addition to coupler and real-ear measures, the target aided sound field threshold (ASFT) is given.

Subject data for an individual with moderate hearing loss given by the MSUv3.1 program.

Prescription for the hearing loss shown in Figure 1.

Desired Sensation Level (DSL 3.1)

The DSL software comes on a single disk. Like the MSU program, hardware requirements are modest: IBM XT, 640 kB RAM, a single doubledensity floppy drive, DOS 3.3, and at least Hercules compatible, CGA video. This was a very common video standard which pre-dated EGA and the newer VGA and SVGA standards. The software should work with almost any monitor. The 137 page DSL manual, "A Computer-Assisted Implementation of the Desired Sensation Level Method for Electroacoustic Selection and Fitting in Children" is clear, well-written, quite comprehensive and covers both the use of the program and much of the theory behind DSL's procedures. The interested reader can find additional information about DSL in other publications (Seewald and Ross, 1988; Seewald, 1992; Seewald et al, 1993a; Seewald, 1994; Moodie et al, 1994). Phone numbers are given for technical support, and different contact people are listed depending upon the type of request, i.e., theoretical development, software/hardware issues, or clinical application. The tutorial information and references should make the manual appealing to faculty teaching individuals first being exposed to hearing aid fitting, such as graduate audiology students. Data is entered by keyboard and manipulated by pull-down menus. Navigation through the program is easy. What is expected of the user is straightforward. The screen displays are easily understood. Help screens are available where the program requests information that might not be intuitively obvious. For operating the program, it was seldom necessary to consult the manual. In addition to the audiometric and prescriptive results, client demographic information can be entered. All relevant information can be stored, retrieved, or printed. A one page report, a hearing aid recommendation form, can also be printed.

As the name of the manual implies, the program is focused upon the amplification needs of children. For example, the speech spectrum used as the input to the hearing aid is a compromise between adult values for males and females and also considers the spectrum of the child's voice when self-monitoring speech (Cornelisse et al, 1991). However, the speech spectrum is not too different from one described by Cox and Moore (1988) that would be applicable to adults. Concerning the size of the external ear, young children tend to have ear canals which are shorter and narrower than adults. A shorter canal will tend to have a REUR peaking at a higher frequency. A smaller sized canal will have a larger RECD. The program calculates prescriptions for young children based upon age specific values for REUR and RECD. This data is given at yearly intervals from birth to 5 years. While the program is specifically geared to these unique aspects of pediatric fittings, the author's opinion is that the software is flexible enough to permit adjustment of key parameters (i.e., REUR and RECD) to make the hearing aid recommendation applicable to older children and adults.

The theory behind the DSL 3.1 procedure is similar to that of MSUv3.1. An approximate bisecting of the dynamic range is the target level (i.e., desired sensation level re: threshold) for the long-term average conversational speech spectrum (LTASS). Speech should remain audible, above threshold, but not exceed the upper limit of the comfort range which then becomes the target RESR. The target RESR is calculated by the program, but it is also listed in a series of tables printed in the manual. RESR increases systematically with increases in hearing loss. For a normal threshold, the RESR varies from 94 to 102 dB SPL, depending upon frequency. For a 40 dB HL hearing loss the RESR increases by about 5 dB over the normal level. Beyond 40 dB HL the RESR increases by about 1 dB for each additional 2 dB of hearing loss. The DSL procedure emphasizes the value of measuring, and representing, all variables with a common reference: in-situ ear canal SPLs. However, all predictions can be made from audiogram thresholds obtained with conventional earphones, so this is the minimum information that needs to be entered into the program. Thresholds can also be entered for measurements calibrated for sound field listening (dB HL), ER-3A insert earphones (dB HL), or RESPLs.

Once thresholds are entered the option is available for supplying additional information about LDLs, REUR, and RECD. The manual describes how the RECD can be obtained. As shown in Figure 3, I have defined custom values for each of these measures. While the program modifies its prescription for changes in REUR and RECD, LDLs are not used to calculate the target RESR. However, a subsequent display allows measured LDLs to be compared to predicted RESRs so that adjustments can be made to the hearing aid, if desired. Figure 4 shows the calculated selection results. Audiogram thresholds are converted to a minimum audible pressure (MAP) which is the RESPL corresponding to threshold. The target DSL and amplified LTASS are given along with the in-situ gain (REAR), insertion gain defining the REIR, and aided sound field thresholds needed to verify the insertion gain. At this point comments can be entered as to the type of hearing aid and earmold. These comments are stored with the patient's file and are printed with the hearing aid recommendation form, but they are not used in any calculations. For example, a Libby horn could be selected, but the REIR would not reflect the expected increase in high frequency gain. Figure 5 shows the target 2-cc coupler use gain and OSPL90 needed to supply the desired REIR and RESR.

Thresholds, LDLs, REUR, and RECD values entered into the DSLv3.1 program.

Selection results calculated by DSLv3.1 for the patient in Figure 3.

Target 2-cc coupler gain and OSPL90 recommended by DSLv3.1 for the patient in Figure 3.

The next step in the process is verification of gain and output. The manual describes how the measures can be obtained for commonly available probe-microphone equipment. Gain can be verified by any one of five procedures: aided sound field thresholds, REIR, REAR, amplified level of LTASS, or 2-cc coupler. The amplified LTASS procedure developed by Hawkins et al (1989) measured the amplified level of frequency modulated (FM) tones adjusted to have a level equivalent to 1/3 octave bands of speech. The two options for output verification are a RESR to a pure-tone sweep at 90 dB, or coupler measures of OSPL90. It is recommended that a RECD be obtained if the coupler method is used. Figure 6 shows the difference between the target and measured gain and output for the two methods that were selected.

Measured versus target gain (REIG) and output (RESR) computed by DSLv3.1.

Figures 7 and ​ 8 are graphs displaying the unaided and aided results associated with the selection and verification procedure (SPLograms). All variables are represented in RESPLs. For each graph, variables can be added to the display sequentially. This should simplify the counseling process if, for example, this graph were to be used to describe the fitting to the parents. First, normal thresholds, the thresholds associated with the hearing loss, and LDL are displayed. This defines the LTLR. By adding the LTASS, it can be seen what part of the speech spectrum is audible and what is inaudible. Adding the amplified LTASS and RESR targets can then portray the effect the hearing aid should have on the speech spectrum. Figure 8 then shows how the measured speech spectrum actually fits into the LTLR, and how the measured RESR compares to the target. Overall, this seems like a convenient tool for taking a structured approach to expressing to the parents the purpose, and success (or failure), of the hearing aid fitting.

Unaided results for the subject data described in Figures 3 to ​ 6.

Aided results for the subject data described in Figures 3 to ​ 6.

The Independent Hearing Aid Fitting Forum (IHAFF) Protocol

Dissatisfied by the lack of procedures and guidelines for fitting more technologically complex hearing aids (i.e., two channel compression, programmable), the IHAFF group met in March of 1993 to begin the process of developing a comprehensive set of protocols for selecting and fitting hearing aids. The IHAFF group is composed of twelve notables in the hearing aid field: Lu Beck, Gail Gudmundsen, Gus Mueller, Ruth Bentler, David Hawkins, Larry Revit, Robyn Cox, Mead Killion, Michael Valente, David Fabry, Michael Marion, and Dennis Van Vliet. The result of their effforts is a set of software tools that run on an entry level PC such as those described for the DSL and MSU procedures. Robyn Cox and associates at the University of Memphis seem primarily responsible for guiding software development, and this university holds the copyright. The program disk is distributed free of charge and contains a comprehensive user manual that can be printed. An overview of the IHAFF protocol can be found in the manual or from Van Vliet (1995). The version described in this paper was 1.1a and the software is still considered to be under development; i.e., new features will be added and modifications will be made. The existing software runs quite well, has an attractive screen layout, and is easy to use. Help screens are available and commands are chosen via pull-down menus which can be accessed either by keyboard or a mouse. Data files can be created, stored, loaded and most of the essential information can be printed. Some graphics screens cannot be printed without a third party program (such as that which comes with the Fig6 software).

Perhaps the most unique feature of the IHAFF protocol is that it addresses the issue of targets for hearing aids that have frequency responses that vary as a function of input level (i.e., frequency-dependent compression hearing aids). Prior to saturation, the performance of linear hearing aids can be described by one frequency response. Compression aids, especially those with circuitry that is activated at low thresholds (kneepoints) must be described by a family of frequency response curves. The target frequency response for any hearing aid wearer may be different for soft, moderate, or high SPL listening levels. Another way of viewing the issue is that, for any given frequency, what is needed is a target input-output function that matches the needs of the listener. Generally, IHAFF is based upon the guideline that, as with most procedures, speech should be audible, but not reach LDL. Also, the aided loudness experience should be comparable to loudness experienced by normal hearing individuals. Speech that normally sounds soft should, when amplified, also sound soft. Intense speech that is normally experienced as loud, should evoke a similar sensation in the hearing aid wearer. The Contour Test and VIOLA (Visual Input-Output Locator Algorithm) are responsible for converting this theoretical concept into the reality of a 2-cc coupler input-output function for a given hearing aid.

The first step in the process is illustrated in Figure 9 where threshold information is entered. Here values have only been entered for the left ear. Thresholds must be entered for insert earphones (like the ER-3A) which have been calibrated on an HA-1 coupler. To allow the program to accurately convert dB HL to coupler SPL, calibration corrections can also be entered. It is believed that the use of insert earphone is preferrable to conventional earphones since the former can be calibrated directly on a 2-cc coupler, the same device used to measure hearing aids. The next step is to perform the Contour Test which utilizes a categorical scaling procedure to measure the individual subject's loudness perceptions. It is assumed that individual variability is too great to use group data on loudness perceptions to predict the performance of the individual, at least with an acceptably low amount of error. It is also recognized that performing loudness measurements can be very time consuming, especially if a great number of frequencies are included. As a compromise (about 1.5 hours estimated for the entire evaluation) loudness measurements are recommended for two frequencies, i.e., a low frequency and a high frequency such as 500 and 3000 Hz. These two frequencies would typically be a good choice for a two-channel programmable aid with a crossover frequency of about 1500 Hz.

Threshold and audiometer calibration information is entered for the IHAFF procedure.

Each frequency is tested by presenting four ascending series of pulsed warble tones (narrow band noise can also be used if the audiogram is not very steeply sloping) encompassing the listener's dynamic range. Step size can be 5, 2.5, or 2 dB. Smaller step sizes are preferred for smaller dynamic ranges. After giving the listener standard instructions, the test begins. At each presentation level the patient chooses which loudness category best describes the stimulus: 0) inaudible, 1) very soft, 2) soft, 3) comfortable, but slightly soft, 4) comfortable, 5) comfortable, but slightly loud, 6) loud, but OK, and 7) uncomfortably loud (Hawkins et al, 1987). The data is evaluated to determine the median SPLs corresponding to each of the categories 1–7. This information can be gathered manually, but IHAFF contains software to drive a set of common, commercially available audiometers equipped with a serial interface. The computer controls the presentation of stimuli by the audiometer, the operator enters the category number of the subject's response, and the software computes the median SPLs. Figure 10 shows the median SPL data, entered manually, for each loudness category. Figure 11 portrays this information graphically and is as close as IHAFF comes to giving a target frequency response. The next step in the procedure involves relating this loudness data to the perception of speech and creating a target to which a hearing aid's input-output function can be matched.

Contour data entered for 500 and 3000 Hz for the left ear.

Threshold and equal loudness contours for the data shown in Figure 10.

Three speech spectra, associated with differing levels of vocal effort, are used to create targets for so-called soft, average, and loud speech. The overall level of the speech is 50, 65 and 85 dB SPL, respectively. SPLs have been measured for 1/3 octave bands of each speech spectrum. Normal hearing subjects have rated the loudness of these bands, and these judgments have been equated to the loudness categories associated with warble tones used in the contour test (Cox et al, 1994a). Categories 1–3 were collapsed to define the soft range, 3–5 define the comfortable range, and categories 5–7 represent the loud range. Consequently, it is known where in the normal dynamic range speech falls. For example, a 2000 Hz warble tone judged equally loud as a 2000 Hz 1/3 octave band of average speech lies at 82% of the distance bounding the soft range. The goal of the hearing aid fitting would be to amplify this same speech signal to the 82% level of the soft range measured for the hearing-impaired listener. Fortunately, algorithms have been developed for expediting this procedure and are incorporated into the VIOLA section of the IHAFF suite (Cox et al, 1994b).

Figure 12 shows the VIOLA screen for the contour data plotted in Figure 11. Two graphs are shown for the frequencies at which loudness data were generated, 500 and 3000 Hz. Output SPL in a 2-cc coupler is plotted as a function of the SPL input to the hearing aid microphone. Corrections are made for type of hearing aid (BTE, ITE, ITC, or CIC). Targets are shown as asterisks representing levels needed to normalize the loudness of soft, average and loud speech. The shaded areas represent the soft (1-3), comfortable (3-5), and loud (5-7) range of the Contour Test. At the top of Figure 12 are a series of boxes where the user can enter parameters which describe the input-output function of the hearing aid: gain at 40 dB SPL, two kneepoints, two compression ratios, and maximum output. The user adjusts these parameters until there is an acceptable match between the input-output function and the targets. Also, the aid should saturate at a level below LDL (the top of the loud range). After the ideal parameters are selected, the program's work is finished. It is the dispenser's job to find a hearing aid with specifications that match. The hearing aid should be adjusted in a test box until the input-output function matches the target. If the aid were to be placed on an average ear, there should be a close approximation to what is needed to normalize loudness for the three speech signals. However, since IHAFF currently has no provision for entering custom REUR or RECD data, there may need to be some final adjustments when the aid is placed on the individual's ear.

VIOLA graph showing possible choices for compression parameters defining the input-output function needed to hit the targets based on the loudness judgment shown in Figures 10 and ​ 11.

The final section of the IHAFF suite is a selfassessment scale, the Abbreviated Profile of Hearing Aid Benefit (APHAB) as described by Cox and Alexander (1995). The APHAB consists of 24 items drawn from the 66-item profile of hearing aid benefit (PHAB) developed by Cox et al (1991) as a research tool. This shortened version takes approximately 10 minutes to administer and the results can be used to determine how well an individual perceives that he or she is doing with and without the hearing aid, thereby deriving an estimate of benefit. There are six questions related to each of four areas: ease of communication (EC), reverberation (RV), background noise (BN), and aversiveness of sounds (AV). The first three scales evaluate speech communication in quiet, reverberant, and noisy situations. The last scale provides input on perception of intense sounds. Responses to the APHAB have been evaluated on a group of 90 men and 38 women with mild-to-moderate, gradually falling hearing loss who wore linear hearing aids. The mean age was 68 years, the range from 30 to 87 years. A similar group of 55 successful (defined as hearing aid use of at least 4 hours per day for at least one year) wearers of linear amplification were also administered the APHAB questions (Cox and Rivera, 1992). Individual performance on the APHAB can be compared to this reference group. Generally, the APHAB has been shown to be a reliable instrument.

The client can take the APHAB by entering responses to the questions via the keyboard (Figure 13). A pencil and paper version can be used for individuals who cannot use the computer, and the clinician can then enter the responses. Once all responses are entered, three graphs are generated. The graph in the upper left corner of Figure 14 summarizes responses for the four subscales during unaided listening. The client's responses are compared to the 20% and 80% percentiles for the successful hearing aid wearers. A similar graph is generated for aided listening (Figure 15), and the resulting benefit is shown in Figure 16. From a statistical standpoint, the 90% critical differences for benefit are 26% (EC), 28% (RV), 27% (BN), and 31% (AV) as reported by Cox and Alexander (1995). Consequently, the results shown in Figure 16 can be interpreted to indicate statistically significant benefit on the EC, RV, and BN scales. When used in this way, the APHAB can be a useful tool for documenting hearing aid benefit.

Screen showing first question of the APHAB when the client enters data.

Display showing unaided results of the APHAB.

Display show aided results of the APHAB.

Display comparing the unaided and unaided results (benefit) of the APHAB.

Ricketts and Bentler (RAB) have developed software which builds upon the IHAFF approach and extends its usefulness; RAB provides level-dependent frequency response targets. Although the RAB software was not available for evaluation, it seemed appropriate to discuss it within the context of IHAFF. RAB is similar in concept to VIOLA in that an attempt is made to restore normal loudness perception (Ricketts, 1996). Research conducted in the development of VIOLA (Cox et al, 1994a) suggests that corrections need to be applied for differences in loudness perception associated with differences in the bandwidth of speech and warble tones. Ricketts and Bentler (in press), however, have found that these differences do not exist provided that the overall SPL of the stimuli are the same and that the bandwidth of the stimuli are less than the critical band, as is the case with 1/3 octave bands of speech and warble tones. RAB estimates hearing aid parameters in two frequency regions, i.e., such as those associated with a two channel hearing aid. Gain in the linear portion of the input-output function is:

LinearGain = (Si − Sn) − (WBf − NBf)

where Si and Sn are the differences, respectively, in signal level needed to elicit a "soft" loudness category rating for the impaired ear and the normal ear. WBf and NBf are SPLs needed to elicit a soft rating for wide band and narrow band stimuli at the frequency tested. The quantity WBf - NBf represents a reduction in gain needed to compensate for loudness summation. Gain at other frequencies is estimated using a procedure similar to the revised National Acoustic Laboratories' NAL-R (Byrne and Dillon, 1986), adjusted by the measured gain. This is an improvement over the IHAFF protocol, which provides no target frequency response other than the two measured frequencies. Compression thresholds are fixed at 48 dB SPL in the low frequency band and 42 dB SPL in the high frequency band. Compression ratios are calculated by the slope of the regression line fitting the targets for soft, comfortable, and loud as would be displayed on the input-output function for VIOLA.

Figure 6 (Fig6) Program

Fig6 is a spreadsheet approach, delivered as an executable program. It is designed to calculate insertion and coupler gain frequency response targets for three input levels (40, 65, and 95 dB SPL). These targets are calculated from the audiogram thresholds alone, and the calculations are based on empirical research involving loudness perception in groups with varying degrees of sensorineural hearing loss (Hellman and Meiselman, 1993; Lyregaard, 1988; Pascoe, 1988; Lippman et al, 1981). The program comes with an 18-page user manual and a number of reprints of articles. Collectively, this material thoroughly describes the theory behind Fig6, formulae for the calculations, and how to install and use the software. The program and related files come on a single 3.5" floppy disk, and they can be automatically installed on a hard drive by executing a batch file called "Setup." Fig6 also contains an online help facility. Patient files, which consist of mostly the patient's name and audiogram, can be saved. All screens can be printed. A second disk, unrelated to Fig6, contains the ELCVIL8 program. This software is based on Villchur's (1973) work and is designed to facilitate selecting parameters for fitting the ReSound programmable hearing aid. The ELCVIL8 program has similar hardware requirements as Fig6.

Killion and Fikret-Pasa (1993) describe three types of sensorineural hearing loss, differing mostly in degree of loss. The Fig6 procedure provides input level dependent frequency response targets intended to normalize loudness for type I and type II hearing loss (Killion, 1996). The type I hearing loss has a magnitude of up to about 40 dB HL and is characterized by a normal inner hair cell population and virtually complete destruction of the outer hair cells. This type of hearing loss requires no gain for more intense signals (80 dB HL and above) because recruitment is complete and loudness perception is normal. There is a loss in loudness for faint signals and gain is needed to compensate for their lack of audibility. A type II loss involves destruction of all outer hair cells and a proportion of the inner hair cells, too. Consequently, with a loss of this magnitude, say 60 dB HL, recruitment will not be complete, even for intense sound exceeding 90 dB HL. However, less gain will be needed at higher levels, than at lower input SPLs, to compensate for loss in loudness. Type III hearing losses are greater than 70 dB HL and require the patient to listen at levels close to the LDL in order to understand speech. The Fig6 manual cautions the user to consider a low distortion aid with compression limiting (variable release time) for hearing losses exceeding 70 dB HL. Figure 6 (from which Fig6 gets its name) of the Killion and Fikret-Pasa (1993) article is a plot of required gain for normal loudness perception as a function of hearing loss. The graph contains three curves; the parameter of the curves is the input level to the hearing aid: 40, 65, and 95 dB SPL. From the graph it is possible to read target insertion gain needed to normalize loudness for soft, moderate and intense inputs, for differing magnitudes of hearing loss. The Fig6 program automatically calculates these targets and uses the appropriate CORFIGs (Table 2 for a definition of CORFIG) to convert insertion gain to coupler gain for BTE, ITE, ITC, or CIC instruments. Figure 17 shows the data entry screen, and a moderate hearing loss has been entered for the right ear. After each threshold has been entered, Fig6 automatically calculates the target insertion gain for that frequency at each of the input levels. Fig6 also calculates compression ratios for soft-moderate (40-65 dB SPL) and moderate-intense levels (65-95 dB SPL) for a low and high frequency band. These compression ratios could be used for programming two-channel instruments. Figure 17 also shows the "Graphs" menu from which the audiogram and various target curves can be displayed.

Data entry screen for the Fig6 program.

Figure 18 illustrates a typical graph showing coupler gain targets (input levels of 40, 65, and 95 dB SPL) for an ITE hearing aid. Graphs for insertion gain, or for other instruments (e.g., BTE or CIC), are similar in appearance. Figure 18 also contains a table which provides information which should facilitate selecting a matrix. The label G40hf , which is 37 dB, gives the average high frequency (hf) gain (G) needed for a 40 dB SPL input. The frequencies used in the average are 3 and 4 kHz. HFE40 is a slope parameter, indicating that there should be a difference of 25 dB between the hf average gain and the gain at 500 Hz for a 40 dB input. Similar information is given for gain for moderate level signals of 65 dB (G65hf) and for intense input levels of 95 dB SPL (G95hf). An additional slope parameter (HFE95) helps to define the saturation response of the aid. The Fig6 procedure produces results very quickly since audiogram thresholds are the only information entered into the program. The targets represent values that should be appropriate for the average listener, because the formulae were developed from group trends. Since individuals vary, some fine tuning of the hearing aid would be expected.

Target coupler gain for 40, 65, and 95 dB inputs calculated by Fig6.

Speech Intelligibility Index (SII)

The speech intelligibility index is a proposed ANSI standard which can be useful for predicting how intelligible amplified speech will be for the hearing-impaired listener. The SII draws upon a considerable body of research validating the technique (Dugal et al, 1980; Pavlovic, 1993; Studebaker and Sherbecoe, 1993), and is based upon an older standard describing procedures for calculating the articulation index (ANSI S3.5-1969). Briefly, both SII and AI calculations result in a number ranging from 0 to 1 (or, if expressed as a percentage, 0 to 100%). This number is a value estimating the proportion of the long term speech spectrum which is audible. Higher numbers are associated with better intelligibility since more of the relevant speech cues are audible. Calculation of the AI takes into account a number of factors: the sensation level of the speech relative to an audiogram threshold or masked threshold, the bandwidth of the speech, the relative importance of different portions of the spectrum to intelligibility, and upward spread of masking. SII calculations additionally include: self-masking of the speech spectrum by adjacent speech bands and a correction for distortion created by high speech presentation levels, an important issue in amplification. Since calculation of the SII is very computationally intensive, it is convenient that a computer program is supplied with the draft standard. The program requires only a PC running DOS 2.1 and 128 kB of RAM. To calculate the SII, the user first creates a text file containing three lines. The first line of the text file contains numerical data about the speech spectrum level, the second line is the equivalent noise spectrum level, and the third line is the equivalent threshold level. In the following examples, a text file is created specifying average conversational level speech, quiet listening (no noise), and a flat 35, 40, and 50 dB HL hearing loss. The first text file is named "normal35.txt," the name of the program is sii.exe, and both the program and data files are in a directory named "ai." Typing

yields the above results, values of 0.54, 0.382, and 0.081 for the SII. The speech intelligibility corresponding to these SIIs depends upon specifics such as characteristics of the speech material (i.e., word or sentence) and talker (male or female), and proficiency of the individual. For W-22 monosyllables, intelligibility would be roughly 95%, 80%, and 5% correct for the SIIs of 0.54, 0.382, and 0.081, respectively (Studebaker and Sherbecoe, 1993). There are limitations to how well absolute intelligibility can be predicted for a particular individual (as opposed to group data). However, it is assumed that higher AI or SII scores predict better intelligibility for the individual. From a practical standpoint, it is quite convenient to use the AI or SII to quickly predict intelligibility for multiple changes in parameters affecting the frequency response. Measuring intelligibility using traditional techniques is extremely time consuming, especially considering the large number of words that must be given to ensure that results are statistically significant (Thornton and Raffin, 1978).

To make calculation of the SII more convenient and more applicable to hearing aid evaluation, the author developed a HyperCard stack called "SII" to implement the 1/3 octave band method for computing SII. Figure 19 shows the card used to enter audiogram information. There are similar cards for entering 2-cc coupler gain, noise spectrum levels, RECD, and REUR. Data are entered using the mouse to drag thresholds to their desired level. The program will calculate intermediate frequencies by interpolating from the octave frequencies. Figure 20 shows the results of the calculations for the audiogram shown in Figure 19. Noise levels were set to a minimum value (quiet) as recommended by ANSI S3.79. HA-1 coupler gain was modified by corrections for an ITE microphone location, REUR was set to normal values for a diffuse field, RECD was normal, and speech was produced with increased vocal effort (68 dB SPL). Four SII values are calculated: SSIa is more similar to the original AI (there is no adjustment for high level speech), SIIb includes the high level speech adjustment (and is the calculation recommended by the standard), SIIc includes a correction proposed by Pavlovic, Studebaker, and Sherbecoe (1986) for additional distortion created by hearing loss, and SIId combines both corrections for speech levels and hearing loss.

An example of aided SII calculations.

Hearing Aid Selection (HAS 2.5)

The current version of the Hearing Aid Selection stack is based upon older programs originally created for the Apple II platform (de Jonge, 1985; de Jonge, 1987). The objective has been to create a structured, scientifically defensible approach to selecting hearing aid characteristics. The program should incorporate relevant findings from the literature (i.e., RECD, microphone placement effects, prescriptive protocols, etc.) and allow the user to apply them in an easy, intuitive manner. The goal was not to espouse any one philosophy but to facilitate the application of different approaches. A primary aim was to introduce graduate audiology students to hearing aid selection, to give them a place to begin. HAS also contains tutorial material. The HyperCard stack uses the Macintosh interface, pull-down menus, buttons, and so on. On-line help and explanations were included to make the "user manual" internal to the program itself, hopefully to improve ease of use. The software and a user manual is available from the Support Syndicate for Audiology (Table 1). HAS can be used to create, save, and print patient files. The files contain all demographic, audiologic, and prescriptive information including a narrative not to exceed 30,000 characters. Target frequency responses, SSPL curves, etc, can also be printed on a single sheet of paper as minigraphs. As with any HyperCard stack, any screen can be printed.

All information in the program is entered for the fifteen 1/3 octave bands contained in the octaves from 250 to 4000 Hz. The highest frequency is about 6000 Hz, corresponding to the high frequency cut-off of the last 1/3 octave band. The audiogram shown in Figure 21 was created by dragging (with the mouse) the thresholds at octave intervals to the desired dB HL calibrated for supra-aural earphones. The program contains default real-ear dial differences (REDDs) for Etymotic ER3-A earphones as well as the TDH-39 headphones. Since the individual REDD can be entered, thresholds are generalizable to any transducer. To speed data entry, the intermediate frequencies were estimated by clicking the "Interpolate" button. In a similar manner, MCLs and LDLs can be entered manually, or they can be predicted. LDLs (labeled "U" for UCL) were calculated using the formula suggested by Pascoe (1988), but LDLs recommended by Seewald et al (1993b) can also be used. In this example, it is assumed that 65 dB SPL, average conversational level (ACL) speech (as defined by ANSI S3.5-1969), will be the input to the listener. The speech level can be raised or lowered covering a range from 45 to 95 dB SPL, or the input level can be adjusted manually, for each 1/3 octave band, to whatever level desired. The effect of the hearing loss on the speech spectrum can be visualized by drawing the audible (lightly shaded) and inaudible (darkly shaded) portion via a menu selection, and the corresponding AI value is calculated and displayed.

A sample screen from the HAS 2.5 program illustrating audiogram thresholds (O), loudness discomfort levels (U), and the audibility of the speech spectrum.

The MCL was estimated using the NAL-R (Byrne and Dillon, 1986) procedure. A variety of other prescriptive procedures are also available: POGO (McCandless and Lyregaard, 1983); bisecting the dynamic range; a 1/3,1/2, 2/3 gain rule adopted from Libby (1986); Berger's procedure (Berger et al, 1979; Berger et al, 1989); DSL 3.1 (Seewald et al, 1993b); or the loudness density normalization (LDN) procedure (Leijon, 1991). Using LDN in conjuction with different input levels, it is possible to generate level dependent targets, such as those given by DSL [i/o]. The program also allows for three custom selection procedures to be used providing the format is that of:

where T is the audiogram threshold, m is a multiplier (e.g., m = 0.5 for a half gain rule) and b is a frequency dependent correction. Figure 22 illustrates the effect upon the ACL speech spectrum of the insertion response specified by NAL-R. The NAL-R REIR is shown by the curve connecting the symbols in Figure 23. Other curves shown in Figure 23 were generated using the LDN procedure to calculate targets for different speech input levels (50, 65, and 90 dB SPL, overall level) for an ear with complete recruitment. Targets can also be calculated for partial recruitment or for recruitment based upon LDLs calculated using Pascoe's (1988) data. Figure 22 also shows the effect of the REIR on speech intelligibility using AI calculations for the unaided (19%) and aided (58%) conditions (ANSI S3.5-1969). Figure 24 shows the program's display of the relationship between AI values and speech intelligibility. One curve is for monosyllables and the other is for sentence material. The added contextual cues provided by sentences make them more intelligible than monosyllables. The data for each curve is stored so that the user can get numerical estimates of the speech intelligibility corresponding to the AI values by selecting a menu item from the cards shown in Figures 21 and ​ 22. Figure 24 also displays critical differences needed to assure that two speech recognition scores are not likely to be due merely to chance. For example, the 58% AI values for amplified speech should create a monosyllable word recognition score of 82% correct. For another word recogniton score to be statistically different it must be greater than 94% or less than 66%, at least for a 50-word list. The program also gives estimates for 10, 25, and 100-item lists (Thornton and Raffin, 1978).

The same audiogram in Figure 21 after the speech spectrum has been amplified by the REIR specified by the NAL-R procedure.

Target REIR needed to produce the amplified speech spectrum shown in Figure 22. Three additonal level-dependent targets are displayed. These target were created by the LDN procedure.

The relationship between articulation index values and speech intelligibility and critical differences.

If one is satisfied with the predicted REIR (the REAR curve can be displayed also) and estimated intelligibility, then the next step is to convert this information to 2-cc coupler terms so that a hearing aid can be found from a manufacturer's specifications (e.g., a matrix). First, the REIR is converted to an estimated 2-cc use-gain curve (not shown), and then the use-gain curve is converted to a full-on gain (FOG) curve by specifying the desired amount of reserve gain. The FOG curve shown in Figure 25 corresponds to 11 dB of reserve gain. So, for an ITE aid, this FOG curve will produce a use-gain creating the target REIR resulting in the previously estimated intelligibility. At this point it is possible to work the procedure in reverse: drag the FOG thresholds to, say, correspond to a particular 2-cc frequency response curve. The FOG will then change the use-gain curve, target REIR curve, and the effect upon the speech spectrum and AI values will be redrawn on the screen shown in Figure 22. Thus, it is possible to visualize the effects of error in selecting a frequency response. Or, the appropriateness of a patient's hearing aid can be estimated.

The HA-1 2-cc full-on gain needed to give the REIR shown in Figure 23.

The LDL curve (Figures 21 and ​ 22) is then converted to a RESR (not shown, but available in the program). Corrections are applied to adjust the real-ear SPL to an HA-1 SSPL90 curve and the result is given in Figure 26. The largest 2-cc coupler ouput corresponding to LDL is estimated to be 116 dB SPL at 3000 Hz. It is then at the user's discretion to decide how close to this curve the output should get. Perhaps a matrix of 110/40/20 would be appropriate given the calculated gain and output. These target responses can be customized for the individual by making adjustments to the default (normal, or average) REUR or RECD curves. Figure 27 gives an example of an adjusted RECD, and changes can be made to the REUR in a similar way. The solid curve shows the RECD used for an average individual (Sachs and Burkhard, 1972). The symbols were dragged to a higher position, reflecting an individual with a larger than normal RECD, as would be expected for someone with a smaller ear canal and higher eardrum impedance. The program will automatically alter 2-cc coupler curves to account for these changes.

Target HA-1 2-cc SSPL90 corresponding to the LDLs shown in Figures 21 and ​ 22.

Card where custom values for the RECD can be entered.

Figure 28 shows target dynamic range compression curves and compression ratios needed to map the dynamic range of the normal ear into the residual dynamic range of the impaired ear for each of the 15 bands. The concept behind this display is similar to that described, in the next section, for DSL [i/o]. By clicking on the "More Info …" button, the user is taken to the card that provides a brief tutorial on the topic of compression and compression ratios (Figure 29). This card is animated to the extent that the user can vary the input level by dragging the rectangle in the normal range, and the rectangle in the reduced range will move according to a 2:1 compression ratio. Figure 29 is shown as just one example of many such tutorials.

Dynamic range compression ratios needed to map the normal dynamic range into the dynamic range shown in Figures 21 and ​ 22.

Tutorial information related to dynamic range compression.

The VIOLA procedure, which is part of the IHAFF suite, and Killion's Fig6 procedure provided the motivation for creating the card illustrated in Figure 30. The input-output function (heavy line) is displayed with the input defined by the location of a reference microphone used during real-ear measures. The output is referenced to the eardrum. Microphone location corrections are being applied for an ITE aid in this example. Microphone corrections are assumed to be level dependent. The full correction is used when the aid is operating in the linear range. When the compressor is activated, the correction is reduced by the compression ratio (e.g., half as great a correction for a 2:1 ratio). No microphone correction is applied when the aid is in saturation. Compression threshold (kneepoints), ratio, and type (input or output) can be defined for two compressors for the currently selected frequency, 1000 Hz in this example. A target input-output function is given based upon Leijon's loudness density normalization procedure (the circles connected by the thin line). Targets (the heavy plus symbols) are also shown for output levels corresponding to 40, 65, and 95 dB SPL inputs. These latter targets (shown as 60, 84, and 104 dB RESPL) are not calculated by the program but are assumed to relate to loudness measurements obtained by another means, such as one similar to the IHAFF contour procedure. Figure 30 also shows a picture of the measurement paradigm, to remind the user of the references for input and output. By dragging the control, indicated by the large triangle under "(dBin)," the input level increases or decreases. Output and headroom numerically change based upon the gain and SSPL selected. A large dot moves along the input-output function to show the user the corresponding point on this curve. Gain and SSPL90 can also be changed by dragging the appropriate triangle. In this example, the targets are not being matched. The gain should be increased somewhat.

Sample input-output functions for a set of compression parameters and the target needed to normalize loudness.

A major concern is how well hearing aids perform in noisy situations. HAS provides two tools for evaluating the effects of noise on speech intelligibility. Figure 31 shows a spectrum level of noise measured in the ear canal with a probe microphone system (the Fonix 6500). Since this system displays the overall level in a 100 Hz bandwidth (BW), subtracting 20 dB from the measured noise converts the measurements to spectrum levels (spectrum level = 10 log(BW)). The masked audiogram thresholds, calculated from these measurements using critical band theory, are shown in Figure 31. The effect of this noise level upon estimated speech intelligibility is illustrated in Figure 32 using AI calculations. The curve indicated by the open circles shows how the AI varies as a function of presention level relative to the input spectrum (which is 65 dB SPL average conversational level speech in this example). The filled circles and squares show the AI function for amplified speech. "Quiet" means that the AI is calculated for the audiogram thresholds reflecting the hearing loss. In this example they are not visible since they are overlaid by the solid squares. "Quiet+Noise" means that the AI is calculated for whichever is larger: the audiogram threshold in quiet or the masked threshold due to noise. "Noise Only" (open squares) means that the AI is calculated based on masked thresholds. For this example, this curve indicates that noise is not a factor in limiting performance. The magnitude of the hearing loss is controlling intelligibility.

An example of calculations showing expected masking effect of noise.

Articulation index functions predicting the effect of noise upon speech intelligibility.

Earmold (or shell) acoustics can have an appreciable effect upon the frequency response of the hearing aid. Adjustments can be made to the frequency response by selecting from a variety of special earmolds, when a BTE aid is selected. Venting can be applied to either BTE, ITE, or ITC styles. Figure 33 shows a typical response for a 1.0 cm long, 0.3 cm diameter parallel vent. The program would modify the REIR and RESR to account for sound bleeding out of the vent and also sound entering the vent.

Desired Sensation Level (DSL 4.0 or DSL[i/o])

The DSL 4.0 software, also called DSL [i/o], has all the features of version 3.1 but in a more user-friendly Windows program. A detailed fifty eight page spiral bound paper manual accompanies the software, and a very comprehensive online help facility makes the program easy to use and understand. There is an extensive references section that integrates background theory with the more practical aspects of the fitting. This should appeal to instructors who would like to encourage students to study the rationale behind the fitting procedure. DSL 4.0 has most of the features of version 3.1. Since those features have been described in a previous section of this manuscript, that discussion won't be duplicated here. The focus will be on improvements or enhancements. Figure 34 shows the "Specification" window, and illustrates what a typical DSL 4.0 screen image looks like. This window displays desired 2-cc coupler SSPL90, gain (10 dB reserve gain), speech levels, and compression ratios for a moderate hearing loss. All commands are selected from the menus shown toward the top of the screen, just below the title bar. Commonly used menu items are represented as icons in the button bar, located just below the menus. Clicking on these icons has the same effect as making a selection from a menu. The status bar, located just below the button bar, contains a series of boxes giving the setting of each of the parameters used in the DSL algorithms. For example, the box containing the text, "Speech: Cox & Moore," indicates that a speech spectrum appropriate to adult listeners has been used for predicting target aided output levels (Cox and Moore, 1988). The extension of DSL 4.0 to adult hearing aid fittings is one of the major enhancements to the software. DSL 4.0 will also allow custom REDDs to be entered, and the software supports a variety of transducers: TDH-style headphones, ER-3A earphones, and sound field loudspeakers.

The "Specification" window from DSL 4.0 showing target OSPL90, gain, and compression ratios.

In addition to linear hearing aids, DSL 4.0 is particularly concerned with fitting wide dynamic range compression (WDRC) instruments. The usefulness of version 3.1 is extended in two important ways: 1) by calculating a family of input-level-dependent target frequency response curves based upon measured thresholds and estimated RESRs and 2) providing target compression ratios and input-output functions. Figure 35 shows the "SPLogram Verification" window for tonal signals measured in the real ear for an ITE aid. Normal thresholds are shown at the bottom of the graph as inverted triangles. The patient's thresholds are given as squares, and the target RESR is the curve at the top of the graph. Between threshold and RESR are a series of curves representing target frequency responses for each of the input levels, 45 to 90 dB. This display can also be shown for target speech levels rather than tones.

The "SPLogram - Verification" window. Target level-dependent frequency response curves are shown.

The compression ratio at each frequency is calculated so that the normal dynamic range is compressed into the hearing-impaired patient's dynamic range. The SPL of sound which is barely audible for the normal hearing listener (normal threshold, Tn) is barely audible for the hearing impaired (impaired threshold, Ti), and sounds which are at the upper limit of comfort for the normal (ULn) are amplified to the upper limit of comfort (ULi) for the hearing-impaired listener (Cornelisse et al, 1994). The compression ratio is then:

CompressionRatio = (ULn − Tn)/(ULi − Ti)

In this formula, compression thresholds (kneepoints) are set very low, corresponding to Tn, but DSL [i/o] permits other compression thresholds to be entered, from these minimum values up to 65 dB. Figure 36 shows a sample input-output function for a 2000 Hz tone. An example of a typical "Help" window is also shown partially overlying the "Input/Output" window.

A sample DSL[i/o] input-output function along with the online help manual and tutorial.

One complication to using hearing aid selection/verification software is the difficulty integrating the software's recommendations with actual hearing aid measurements, either real ear or coupler. For example, if a target frequency response curve (or input-output function) is calculated, it would be very useful for that target to be displayed on the equipment used to measure the hearing aid. The Audioscan helps to overcome some of these difficulties. Software version 2.6 of the Audioscan hearing aid test system (Etymonic Design, Inc.) implements a subset of the DSL 4.0 program. Audiometric data can be entered and target values for RESPL are calculated for ACL speech and also for the RESR. While the Audioscan does not calculate target level-dependent frequency response curves, the screen does show amplified ouput levels for soft, average, and loud speech. This display is not shown, but it is similar to the SPLograms that in Figures 7 and ​ 8. The user can easily visualize whether soft sounds are above threshold, if ACL speech is hitting the target frequency response, and whether loud speech is below the RESR. The Audioscan does not give a target input-output function, but a display is shown of how well the hearing aid compresses the speech signal into the residual dynamic range. These displays can represent the actual real-ear output for the individual, or the output can be simulated from 2-cc coupler measures. The hearing aid can be adjusted on the 2-cc coupler, and then the fitting can be verified on the individual. This is a very convenient approach with patients who don't readily cooperate, such as children. The simulation can be made more accurate by taking into account the patient's RECD and the REDD.

Hearing Aid Selection Program/Procedure (HASP 2.07)

The emphasis of the hearing aid selection programs previously discussed has been upon identifying appropriate prescriptions. These prescriptions usually account for differences in REIG and coupler gain, but often they do not account for venting or earmold acoustics. The HAS program does make predictions concerning venting and earmold effects, but it is still left to the dispenser to find an actual hearing aid that will meet the prescription. Here is one area where HASP departs significantly from the previous approaches (Dillon et al, 1992). The program comes with a small core database of hearing aids and, additionally, two large databases of hearing aids from manufacturers (Miracle Ear and Starkey). Once a target prescription is identified, the computer will automatically search the databases to find hearing aids which match. These databases were created by the manufacturers so the dispenser is spared the tedious chore of entering this information manually. However, if desired, the dispenser does have the option of including hearing aids from other manufacturers. It is recommended that HASP run on a PC AT compatible computer using a color monitor and a minimum of 384 kB of RAM memory. The HASP suite of programs and database files comes on a single disk and should be installed on a hard drive. This requires 0.7 MB of space. The fifty one page manual is attractively bound, has an index, references section, description of the theory behind the approach, along with detailed instructions for using the program.

HASP calculates target insertion gain curves using the NAL-R procedure (Byrne and Dillon, 1986; Byrne et al, 1990). The minimum SSPL90 is selected so that 70 dB SPL speech will not saturate the hearing aid. The maximum SSPL90 should not exceed LDL (which is estimated from threshold). The target SSPL90 is the midpoint of these two curves. HASP allows the user to select vent size, earmold configurations and dampers for BTE instruments. The computer adjusts the frequency response and SSPL90 for sound entering and exiting the vent (Dillon, 1991), considers the occlusion effect, and also can predict the maximum gain available before feedback will occur. The computer "understands" the effect of tone controls and output limiters for the hearing aids within its database, and the program will search for aids which achieve an acceptable match. HASP calculates the root-mean-square deviation, adjusted by AI weights, between the target and the hearing aid's response. Acceptable matches can be ranked for "closeness of fit."

Figure 37 shows the results of an automatic search when an ITE hearing aid was requested for a moderate hearing loss. The required REIG (RREG) and required aided sound field thresholds (RAT) are displayed along with the audiogram (HTL). HASP identified eleven possible hearing aids and displayed a variety of useful information. Vdev estimates the dB deviation of the volume control from a mid rotation, COF (closeness of fit) is the rms deviation in dB of the REIR from the target response, and FBM (feedback margin) indicates the number of dB of additional gain available before feedback will occur. HASP can also be used in a manual mode so the dispenser can visualize the effects of modifying the hearing aid. A particular hearing aid is selected from the database, its parameters (such as venting, dampers, tone control) are defined, and HASP displays the expected REIG (EREG) and the amount of deviation from the target (Figure 38). HASP also has a menu option for "coupler gain mode" so that a prescription can be calculated. Here the required coupler gain (RCG) needed to match the REIR target is displayed for a particular style of hearing aid, venting, and tubing options (Figure 39). The RCG is useful for adjusting the hearing aid in the test box prior to fitting the client and performing the real-ear verification. Since HASP has no mechanism for entering custom REUR or RECD values, real-ear verification is an important aspect of the fitting process. RCG is also useful if real-ear measures cannot be performed.

A sample HASP screen showing the results of an automatic search for an ITE aid.

HASP display illustrating information available when used in the manual mode.

Coupler gain required to match the target REIR.

The database stores relevant information about the electroacoustic characteristics of the hearing aid. The computer, however, has no knowledge about issues that are important to the fitting, but which are difficult to quantify: manipulation skills of the client, years of hearing aid use, and other special needs that the client may have. The dispenser can then combine this factual information with clinical judgment and experience to provide the best instrument. In this way the human is left to do what he or she does best, while the computer can handle the chore of calculating corrections and remembering all the details of the hearing instrument inventory.

SUMMARY

An issue, and perhaps a limitation, which is common to all the software packages reviewed is: How well does the program fit into the dispenser's practice? Ideally the programs would be integrated so that, for example, the patient's name and address would not have to entered repeatedly into different software modules. An audiogram entered once would not have to be duplicated. An audiogram stored in a computer file would not also have to be stored on paper. Tympanogram information or other diagnostic tests (e.g., auditory brainstem response, word recognition scores) would be stored within the same patient record. Ideally, the software would be part of the hearing aid test systems so that real ear and coupler measures could be easily compared to the targets. The dots (representing loudness calculations) from VIOLA would appear on the screen of the test box running the hearing aid's input-output function. The same software used to select ideal characteristics of the hearing aid would also be used to adjust the aid, as in the case of a programmable. Or, for conventional hearing aids, a direct link to the manufacturer's database would be useful for matching a matrix. It would also be useful, when trying to match a target coupler gain curve, to simulate the effect of making an adjustment to a trim pot. Then, the hearing aid could be ordered online, and billing would be facilitated by the software. Perhaps the NOAH software platform, together with cooperation from hearing aid manufactures and software developers, will help make such high level integration a practical reality.

While improvements will surely come, a variety of hearing aid selection programs are currently available. Each program has its own set of features, philosophy, and is geared toward different aspects of the fitting process. Some of the procedures are more abstract and theoretical, others are practical and concrete. Certain procedures emphasize the prediction of hearing aid characteristics from threshold data, while others require measurement of loudness perceptions. Hearing aid dispensers also have preferences and a perspective shaped by their personality and experiences. This dictates what they value. Some perceive the hearing aid fitting to be a conceptual experience to be researched or taught to students, others see it as a social experience shared with the patient for whom they care. Others may visualize it as an integral part of the business they have labored to create. Regardless of the perspective, each person can find something to appreciate in the software that has been reviewed. Powerful, inexpensive microcomputers are readily available. A minimal amount of time (less than a few minutes) is often needed to enter data and receive prescriptions that incorporate decades of research. Patients have more confidence in the opinions of their dispensers when they can see the reasoning that went into the recommendations. Allowing patients to watch the computer screen, even though they cannot appreciate most of the details, is an engaging process and allows them to feel they are an active participant. Hopefully, readers will be encouraged to try some of these software tools to add more science and objectivity to the art of hearing aid fitting.

ACKNOWLEDGEMENTS

I would like to thank Robyn Cox, Richard Seewald, Mead Killion, Toni Gitles, and Harvey Dillon for sending software to assist with the review. Ruth Bentler provided details concerning the research she and Todd Ricketts undertook in creating the RAB procedure. Dennis Van Vliet clarified some details relating to IHAFF. Larry Humes and Joe Smaldino helped to make me aware of the availability and changes in software.

Computer Programable Hearing Aid

Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4172261/

Share this post