The Prescription Of Spectral IQ - StarkeyPro

The Prescription of Spectral iQ*Spectral iQ is a frequency lowering algorithmintended for use in cases of severe-to-profound,high-frequency, sloping hearing losses. This noveltechnique for frequency lowering is technicallydescribed as Spectral Envelope Warping (SEW)(Fitz et al., 2009). The SEW algorithm operates bymonitoring high-frequency regions for spectralpeaks that are responsible for the recognition ofhigh-frequency speech sounds like /s/ and /sh/.The algorithm characterizes the high-frequencyspectral shape and dynamically recreates thatspectral shape inside a lower frequency targetregion. The source region for analysis, targetregion for the newly created spectral cue and gainapplied to the newly generated cue are eachprescribed by the Inspire programming software.Candidates for Spectral iQ are those patients whomeet the following audiometric criteria: A ll thresholds below 1000 Hz must be55dBHL or better H igh-frequency hearing loss slope must begreater than or equal to 25dBHL per octave A single threshold between 1000 Hz and3000 Hz must be 55dBHL or worse A ll thresholds between 4000 Hz and 8000 Hzmust be 55dBHL or worseSpectral iQ will automatically activate for patientswho meet these audiometric criteria. Inspirereveals two Spectral iQ controls: Spectral iQBandwidth and Spectral iQ Gain. There are sevenoptions for Spectral iQ Bandwidth and 10, 1dBsteps for Spectral iQ Gain. At default settings,Spectral iQ Bandwidth is set to 3 or 5 (dependingon severity of loss) and Gain is set to 3dB. Each ofthese is described below.The prescription of Spectral iQ is done on anindividual audiometric basis; different patients willbe prescribed different algorithm parameters.Figure 1 illustrates variable parameters that areaffected by Spectral iQ Bandwidth adjustments.Two audiometric markers (Figure 1, panel A,shown with red arrows) are used to define upperand lower frequency boundaries inside whichseven target regions (Figure 1, highlighted inyellow) are defined. Each target regioncorresponds to a Spectral iQ Bandwidth setting inInspire. A target region is the range of frequenciesinside which lowered cues are presented. Thelowest frequency boundary possible for a targetregion is the corner frequency: this audiometricmarker is the lowest frequency that starts the20dB per-octave audiometric slope. In the case ofaudiograms that do not exceed this slope, thelowest frequency that starts the audiometric slopewill be defined as the corner frequency. Thehighest frequency boundary possible for a targetregion is the 70dBHL frequency: the exactfrequency for this 70dBHL threshold is determinedthrough linear interpolation between themeasured audiometric frequencies. IncreasingSpectral iQ Bandwidth (i.e. increasing sensitivity)generates target regions that fall closer to thelow-frequency boundary, or corner frequency;lower Bandwidth settings (i.e. lower sensitivity)generate target regions closer to the highfrequency boundary or 70dBHL frequency.In all panels of Figure 1, a sample patient’saudiogram is shown, the source region for analysisis shown in green and the target region is shown inyellow. In panel A, Spectral iQ Bandwidth isprescribed to more aggressive parameters. Thesource region for analysis covers a broad range ofhigh frequencies, while the target region fallsbetween 500 Hz and 2000 Hz. The example inPanel A may correspond with a Spectral iQBandwidth setting of six or seven. In panel B amoderate setting is illustrated. The source regionThe Prescription of Spectral IQ1

ATarget RegionSource Region020Threshold (dB HL)for analysis remains the same and the targetregion has narrowed in frequency. Panel C showsan illustration of mild Spectral iQ settings. Thesource region for analysis has moved higher infrequency, focusing only on the highest frequencyinformation and the target region has narrowedand shifted higher in frequency.406080100120140100The source region (Figure 1, highlighted in green)for analysis changes with different prescriptions.At less aggressive settings, the source region foranalysis is constrained to only the highestfrequencies; at more aggressive settings, thesource region for analysis will extend into lowerfrequencies.Verification andValidation of Spectral iQ100010000Frequency (Hz)70 dBHLThresholdCornerFrequencyBTarget RegionSource Region0Threshold (dB HL)20406080100120140100100010000Frequency (Hz)CTarget RegionSource Region020Threshold (dB HL)Spectral iQ Gain refers to gain applied to the newlygenerated spectral cue. This value is relative togain prescribed across the target region. Forexample, if the target region spans 1000 Hz and3000 Hz and the gain applied in that range is 20dB,a Spectral iQ gain setting of 3dB will causeSpectral iQ to generate new cues that peak at23dB inside the target region.406080100120140100In the fitting of hearing aids, it is routine clinicalpractice to take a measurement of hearing aidoutput in a patient’s ear. This process of real-earmeasurement (REM) verifies sound levels for thefitted hearing aid and allows comparison to apatient’s audiometric threshold and prescriptivetargets. Because Spectral iQ dynamically capturesand recreates high-frequency spectral cues in realtime, the averaging of level that occurs incommercial real-ear systems makes traditionalreal-ear measurements an inaccuraterepresentation of the feature. However, real-earmeasurements can be used to verify the activity ofSpectral iQ. By comparing measurements withSpectral iQ ON and OFF, an increase in levelshould be observed in the target region withSpectral iQ active. Figure 2 shows two real-earaided responses measured using an AudioscanVerifit and a speech sample that includes a100010000Frequency (Hz)Figure 1. An illustration of three Spectral iQprescriptions are shown at different Bandwidthsettings. The audiogram in each panel isrepresentative of a Spectral iQ candidate. Greenillustrates the source region for analysis and yellowthe target region. Panels A, B, and C provideillustrations of Spectral iQ Bandwidth at Maximum,Moderate and Minimum settings, respectively.band-pass filter centered at 6300 Hz. The greenresponse shows Spectral iQ OFF and the magentaresponse shows Spectral iQ ON. This differencebetween 2000 and 4000 Hz clearly demonstratesthe frequency lowering behavior of Spectral iQ.However, it is important to note that the measureddifference between the time-averaged, real-earcurves and the proximity of the Spectral iQThe Prescription of Spectral IQ2

ReferencesFitz, K., Edwards, B., & Baskent, D. (2009). U.S. Patent Application.20090226016. Washington, DC: U.S. Patent and Trademark Office.Robinson, J.D., Baer, T., & Moore, B.C. (2007). Using transposition toimprove consonant discrimination and detection for listeners withsevere high-frequency hearing loss. International Journal of Audiology,46, 293-308.Please contact jason galster@starkey.comwith questions.Figure 2. Two real-ear aided responses (dB SPL)are shown as a function of frequency (Hz). Themeasurement stimulus (included in the AudioscanVerifit) was a speech sample with high-frequencycomponents band–pass filtered around a centerfrequency of 6300 Hz. The green response showsSpectral iQ OFF and the magenta response showsSpectral iQ ON.response to the audiogram are not expected tocorrelate with benefit from the feature. Thestimulus presented by Spectral iQ will be averagedover time to appear lower in level during real-earmeasurement; meaning that slight differences inthe measured real-ear response may representlarge changes in quality and audibility for thepatient.The recommended process for validation ofSpectral iQ is through the use of high-frequencyspeech sounds. In-office presentation of an S-test,originally described by Robinson, Baer, and Moore(2007), can be used to document a patient’s abilityto detect high-frequency speech sounds. TheS-test is a task of English language pluraldetection that scores correct detection of wordfinal /s/ and /z/. The University of Western OntarioPlurals Test (Glista & Scollie, 2012) is an option fora standardized, recorded test corpus. A worksheetis included at the end of this document that can beused to facilitate an in-office, live-voice S-test.* Spectral iQ is a Starkey brand name.The Prescription of Spectral IQ3

Plural Word Lists for the Validation of Frequency LoweringInstructions for live voice validationWith frequency lowering deactivated, complete the following steps:1. With mouth obscured, say a single word at a conversational level.2. Ask the patient to repeat the word.3. Mark the response as correct or incorrect, completing a single list.4. Calculate the patient’s score (% correct).With frequency lowering activated, use an alternate list and repeat steps 1-4. If the patient’s ability to detectthe plural cues does not improve, increase Spectral iQ Bandwidth and Gain. Continue this process until thepatient demonstrates an improvement in performance with Spectral iQ active.LIST 1SCORELIST 2SCORELIST e Prescription of Spectral IQ4

LIST 4SCORELIST 5SCORELIST CORENotesThe Prescription of Spectral IQ5

ing 2015 Starkey Hearing Technologies. All Rights Reserved. 85441-007 9/15 WTPR0022-01-EE-SG

Spectral iQ Gain refers to gain applied to the newly generated spectral cue. This value is relative to gain prescribed across the target region. For example, if the target region spans 1000 Hz and 3000 Hz and the gain applied in that range is 20dB, a Spectral iQ gain setting of 3dB will cause Spectral iQ to generate new cues that peak at