DSL for Bone Anchored Hearing Devices: Prescriptive Targets and Verification Solutions

Prescription and verification help us achieve consistent and beneficial levels of audibility across patients with air conduction hearing aids (McCreery, Bentler, & Roush, 2013; McCreery, Brennan, Walker, & Spratford, 2017; Moodie, Scollie, Bagatto, & Keene, 2017). Prescriptions are designed for use with nonlinear signal processing such as wide dynamic range compression. Modern air conduction prescriptions can work accurately with nonlinear devices because they are built on well-defined and realistic speech signals (Holube, Fredelake, Vlmaing, & Kollmeier, 2010) and use speech-based nonlinear input/output target functions (Scollie et al., 2005). We verify and fine tune to match prescriptive targets with high quality, speech-based test signals. Verification measurements capture the aided speech spectrum, and the range of aided speech as percentiles (sometimes called peaks and valleys). Using these tools, the aided levels of speech displayed on the verification system are valid measurements, regardless of whether or not the hearing aids are linear or nonlinear. In recent years, bone conduction devices moved to more nonlinear signal processing. Therefore, as with air conduction devices, it may be better to adjust and fit bone conduction hearing devices with speech-based verification (rather than with aided warble-tone thresholds). A 2010 study compared aided thresholds versus suprathreshold aided speech verification with bone conduction hearing devices. The results indicated that it's more fruitful to use suprathreshold verification of aided speech, if we are interested in measuring device performance for realistic input levels (Hodgetts Hagler, Håkansson, & Soli, 2010). In summary, one thing that air conduction and bone conduction hearing aids have in common is that assessing their performance for speech is best done with speech input test signals.

High quality hearing aid fittings that provide audibility of a broad bandwidth provide access to specific speech sounds. Two recent studies using air conduction hearing aids highlight this point (McCreery, Brennan, Walker, & Spratford, 2017; Scollie et al., 2016a). McCreery and colleagues (2017) looked at children's hearing aid fitting data to see if deviations from prescriptive targets impacted speech recognition in quiet and in noise. In other words, if children have good fittings, particularly in the high frequencies, does it affect speech recognition? They found that children who had lower sensation levels, and therefore poorer audibility of speech at 4 kHz, also had poorer speech recognition in quiet and poorer speech recognition in noise. The results indicated that children whose hearing aids were fitted to provide audibility across a wide range of inputs and frequencies had better speech recognition outcomes. Clinically, this means that verification and fine tuning of frequency response shape may be an important precursor to hearing aid benefit. A related idea is the use of specific speech sounds to crosscheck the fitting. For instance, we have developed methods for evaluating frequency lowering signal processing that include verification with a simulation of the speech sound "s" (Scollie et al., 2016a). This allows us to look at whether or not the high frequency component of the hearing aid fitting is providing access to that particular speech sound. These studies highlight the rationale for having hearing aid verification tools to assess the frequency responses of fittings for speech. In fitting air conduction hearing aids, the DSL Method displays the prescriptive targets, hearing aid output, and patient thresholds on one display called the SPLogram to facilitate interpretation of audibility.

There is a reason to use our verification approaches to compare and document the impact of signal processing and the effectiveness of accessories. In one study, we observed that the activation time and the activation strength of noise reduction systems varied significantly across brands and settings in commercial hearing aids (Scollie et al., 2016b). Because of this, in our current protocols we emphasize the importance of the fitting audiologist knowing the activation speed and strength of the processor. Similarly, feedback managers are not all created equal. Some feedback managers affect the frequency response and others do not. Remote microphone and streaming devices can have connection issues or breakdown – without electroacoustic measurements it can be challenging to diagnose or document these problems. We know that hearing aid analyzers provide good objective solutions for addressing these particular technologies, but they've only been available for use with air conduction hearing aids until recently.

There has been a lot of evolution, change and development in bone conduction devices in recent years. There are four families of bone conduction devices as shown in Figure 1: those that are worn and coupled to an abutment (percutaneous bone anchored hearing aid (BAHA); also referred to as percutaneous bone anchored hearing devices (BAHD); those that are fully implanted (active transcutaneous implanted transducer); those that send the signal through the skin with a headband system (passive non-magnetic attachment); and, those that use an implanted magnet (passive transcutaneous implanted magnet). Today's presentation focuses on percutaneous bone anchored hearing devices (BAHD). Percutaneous devices are those that are worn on a surgically implanted, bone-anchored abutment. There are several different brands and models of devices that fall into this particular category.

Many clinicians are familiar with the probe tube microphone systems and air conduction coupler microphone sets that are used for measuring the output of air conduction hearing aids. For measurement of bone anchored hearing devices, an invention called a skull simulator has been available for some time. A research-grade skull simulator (TU-1000) was developed by Håkansson and Carlsson in 1989, and has been used extensively, in research studies. It wasn't designed or commercially produced as a clinical tool, but it has been used in research labs and in industry. The skull simulator is analogous to a coupler microphone. The bone anchored hearing device is attached to the abutment of the skull simulator. The bone anchored device vibrates in response to sound and delivers the force output that it generates into the skull simulator, which is connected to a measurement device. This allows measurement of the aided output of the bone anchored hearing aid, much like making a coupler measurement for air conduction hearing aids.  

Additionally, an on-head microphone is in early development (Hodgetts, Scott, Dylan, Maas, & Westover, 2018).  It is at a research stage right now, so it is not yet available for clinical use. It is analogous to a probe microphone. With this on-head microphone, it may be possible to verify levels of aided speech or other signals from a wider variety of bone conduction hearing systems.

Verification Lessons from the Lab

One thing that we have learned is that just like with air conduction hearing aids, audibility and bandwidth matter. Earlier, we reviewed research by McCreery et al. (2017) revealing that the actual levels of audibility and the bandwidth fitted via an air conduction hearing aid affected the benefit from those devices. Hodgetts and colleagues found similar results using abutment-worn, bone conduction devices (Hodgetts, Hagler, Håkansson, & Soli, 2011).  In their study, each participant performed speech recognition tasks using two different bone anchored fittings. They measured the output of the abutment-worn device as a function of frequency. They plotted the patient's thresholds alongside the verified results. This type of display is similar to the SPLogram that we use in fitting air conduction hearing aids. In the study, they provided one fitting type that had an aided speech spectrum that peaked in the mid-frequencies, and rolled off in the low frequencies and high frequencies. The alternative fitting used a research device to provide a flatter, broader frequency response with more high frequency gain and improved high frequency audibility. 

The participants performed speech recognition tasks with both fittings, and had better results with the fitting that provided better high frequency audibility. This study indicates that for users of bone anchored hearing devices, optimization of benefit is about more than just providing the hearing devices. Just as with air conduction hearing aids, it is important to fit a device that has been fine-tuned, adjusted, and verified to provide that individual with an appropriate and optimized amount of audibility across frequencies. That really comes down to prescription and verification.

Just like with air conduction hearing aids, bone conduction devices have a different response when worn than they do on the coupler. With air conduction hearing aids, the real-ear-to-coupler difference (RECD) is measured across frequencies and represents the difference of the output in the ear versus in the coupler.  From many normative studies we know that earmold RECDs have a mid-frequency boost compared to foam-tip RECDs, and we know how this varies across ages. We use this knowledge in our prescription and verification as a transform, especially to generate age-predicted RECDs when measured values are not available (Bagatto et al., 2005). A similar issue exists for bone anchored hearing devices. With bone anchored hearing devices, there is a real-head-to-coupler difference (RHCD) (Hodgetts et al., 2018). The RHCD is the difference between the device's response when it is worn on the patient’s abutment versus when it is on the skull simulator. It has a different shape than an RECD, but conceptually is similar to an RECD. In the prescription and verification method discussed below, the RHCD is equivalent to an RECD and is used in similar ways.

This defines a series of steps to follow in clinical practice, and protocols that define key measurements to make. Step one is assessing the hearing for the patient via their abutment. Using the patient's own device, we measure the direct bone-conduction thresholds so that they incorporate how the device responds on the head of the individual patient. These in-situ thresholds are measured as dial-levels measured via the device-specific programming module. The dial levels then get converted through the DSL prescriptive software to force levels on the abutment.  

In step two, we compute prescribed targets for speech that are recommended listening levels for that particular patient. The targets will be shaped across frequencies, compressed across input levels, and will be limited to the device-specific maximum force output. In step three, we verify the output of the device (e.g., via a skull simulator), using a speech signal as the input. We assess the fit to targets on the FLogram. We fine-tune the device as needed to optimize the fit to target, and interpret the audibility of the hearing aid fitting by comparing the aided speech frequency responses to the force level thresholds of the device user.

This clinical workflow is similar to the workflow used with air conduction devices.  The key differences with bone conduction devices are: (1) we measure the thresholds in-situ with the device on the abutment using the programming module, and (2) we verify the device when connected to a skull simulator.

First Version of DSL-BCD Targets

In developing the DSL-BCD method, we based the first version of DSL-BCD targets on DSL v5. We started by only applying the transforms necessary to convert DSL v5 from dB SPL to dB FL. This meant that the target shaping and audibility goals were the same as for air conduction hearing aids, with the exception of the loudness and dynamic range adjustments described above. An example of this early target is shown in Figure 6. The targets used wide dynamic range compression, mainly to overcome the limited output dynamic range of the devices. Similar to DSL v5, the DSL-BCD version 1 targets had more gain and output for children; frequency shaping for audibility; a slight reduction for binaural fittings in adults; and targets that fall below threshold for very low input levels, especially if they are below compression thresholds. The targets are, in many ways, similar to DSL air conduction targets.

We completed a validation study to assess the DSL-BCD v1 target. We compared the targets to the actual levels of aided speech used by 39 successful percutaneous bone-anchored device users with a wide range of hearing levels. These adults were fitted at the iRSM clinic in Edmonton, Alberta, Canada. The fittings were completed in a clinic that has a large patient case load that specializes only in fitting this device category, and that monitors the outcomes of their patients.

They were regular wearers of their devices, using them for at least eight hours per day for at least two months, so they were highly acclimatized listeners. Each user’s hearing aid was measured at their own user settings. Measures were made using running speech as an input, for a test level of 65 dB SPL. We used the ISTS signal (Holube et al., 2010) and filtered it to include microphone location effects appropriate for bone-anchored hearing aids, on a research skull simulator. We compared the measured output to the DSL-BCD v1 targets for each patient.

In the high frequencies, we found that high frequency targets and fittings were not significantly different. In the mid-frequencies, the measured device outputs and targets were fairly close, with a 4 dB difference on average. This difference showing mid-frequency targets to be slightly higher than the typically worn values was attributed to an unwanted resonance in the device that could not be resolved with fine tuning. However, in the low frequencies, targets were 19 dB higher than where the successful users actually wore their devices. This was attributed to a necessary fine-tuning change to fit the devices to the user’s preferences.

Considering all of these results, low frequency targets required further adjustments. An adjustment was therefore built into the frequency shaping of the DSL-BCD algorithm to better approximate the preferred frequency response of these users, while the audibility of the mid- and high-frequency targets was maintained at the originally prescribed levels. In summary, the difference between DSL-BCD v1.0, which was the preliminary version, and v1.1, which is the version that is available now, is this low-frequency adjustment. The modified v1.1 target agrees with the fittings of the 39 patients within less than two dB across frequencies.

We also evaluated the aided Speech Intelligibility Index (SII) values for these fittings. These data indicate that these users had high and consistent SII values that vary with the degree of hearing loss. The SII values vary between 75% to about 95% when the four-frequency pure-tone average is about 60 dB or less, and decline as the hearing losses exceed this level. This overall pattern of SII varying with degree of hearing loss is consistent with SII trends with air conduction fittings (Moodie et al., 2017).

Applying DSL-BCD v1.1 in Clinic with the Audioscan Verifit Skull Simulator

As a practical tip, remember to ensure that the air conduction microphones in the test box have been calibrated before moving to the skull simulator. This step is necessary to ensure that the reference microphone(s) in the test box, which will still be used for bone conduction device testing, are calibrated and functioning correctly. After completing this step, unplug the air conduction coupler microphone(s) and plug the skull simulator into the coupler microphone jack in the test box (Figures 8 and 9), shown for both Verifit2 (VF-2) and Verifit1 (VF-1).

It's important to position the skull simulator correctly. Position the skull simulator sideways to the test box speaker, so that the front facing microphone on the bone conduction device is directly facing the speaker. Depending on the device side you are testing, orient the skull simulator so that the red marking is facing you when verifying right devices and the blue marking is facing you when verifying left devices. This placement procedure is shown in Figure 8 and can also be found in photos located within the ‘Help’ system integrated into the Verifit. Note that with the Verifit1 system, the skull simulator is too tall to allow the lid to close, so running tests in a quiet room is essential. After set up, you can run quality control tests, coupler tests, manual control tests, or adjust fittings to targets.

Figure 10 shows the menu that appears after you've selected bone anchored hearing devices (BAHD) as the hearing aid type. Selecting "BAHD" indicates to the software to move into bone conduction fitting mode, and to use the skull simulator for measurement rather than the air conduction coupler microphones. At that point, you will see DSL-BCD for the bone-conduction prescription in the upper left corner of the fitting screen, so you know that you're working in the skull simulator verification mode. 

Case 1 is a 65-year-old male, with a history of bilateral chronic ear disease with multiple tubes and bilateral mastoidectomies. This patient became a candidate for an abutment-worn solution, and was implanted in 2013. He has been a long-standing user of abutment-worn hearing devices. His audiometric information is shown in Figure 11, however, we did retest his hearing by direct bone conduction using the hearing aid fitting software and the device worn in-situ. This process tests through the abutment to remove any effects of conduction through skin and other tissue. The direct thresholds, not the thresholds from the audiogram, are the ones entered into the software.

We would customize the targets for this case as shown in Figure 12. We chose DSL-BCD-adult rather than the child prescription, based on the patient's age. Next, we selected the device make and model that we are fitting. The patient had recently updated to a new device, and that was the reason for the refitting. Remember, selecting the device make and model defines the maximum output target. On the left in Figure 13 is the FLogram. The line closest to the bottom is a normal set of thresholds that is equivalent to zero dB HL, but as measured through an abutment. This patient has a slight threshold elevation relative to that, as depicted by the red line. The patient’s targets are shown in Figure 13 for maximum output (orange symbols) and for three levels of speech ranging from 55 through 75 dB SPL (green, pink, and blue symbols).

Case 2 is an eight-year-old girl with unilateral atresia and microtia. She has been a successful softband user, and she recently received an implant and an abutment-worn device. She has normal hearing thresholds by direct bone conduction, and normal hearing thresholds by an audiometric bone oscillator, as shown in Figure 16.

This patient’s direct bone-conduction thresholds were entered into the fitting system. We customized the targets to be child targets (DSL-BCD child), and we selected the specific hearing aid make and model that she was using so that the maximum force level targets would be appropriate for her fitting. Her previously-worn bone-conduction hearing aid was convertible for use on her abutment, so the main purpose of this visit was to re-adjust the device for use via direct bone conduction.

Her fitting results are shown in Figure 17. We were able to meet targets for maximum output and for three levels of speech. In this case, the aided SII values range between 87 and 94%, which is similar to the SII that we would expect for someone with normal hearing or an aided mild hearing loss, if hearing by air conduction. These values are somewhat higher than the aided SII values we observed for the adult case discussed previously, which is consistent with adult-child differences in SII for air conduction hearing aid fittings.