Swept-tone stimulus-frequency otoacoustic emissions: Normative data and methodological considerations

Abstract

Stimulus-frequency otoacoustic emissions (SFOAEs) are reflection-source emissions, and are the least familiar and perhaps most underutilized otoacoustic emission. Here, normative SFOAE data are presented from a large group of 48 young adults at probe levels from 20 to 60 dB sound pressure level (SPL) across a four-octave frequency range to characterize the typical SFOAE and describe recent methodological advances that have made its measurement more efficient. In young-adult ears, SFOAE levels peaked in the low-to-mid frequencies at mean levels of ∼6-7 dB SPL while signal-to-noise ranged from 23 to 34 dB SPL and test-retest reliability was ±4 dB for 90% of the SFOAE data. On average, females had ∼2.5 dB higher SFOAE levels than males. SFOAE input/output functions showed near linear growth at low levels and a compression threshold averaging 35 dB SPL across frequency. SFOAE phase accumulated ∼32-36 cycles across four octaves on average, and showed level effects when converted to group delay: low-level probes produced longer SFOAE delays. A "break" in the normalized SFOAE delay was observed at 1.1 kHz on average, elucidating the location of the putative apical-basal transition. Technical innovations such as the concurrent sweeping of multiple frequency segments, post hoc suppressor decontamination, and a post hoc artifact-rejection technique were tested.

FIG. 1.

SFOAE amplitude spectra from 45 young-adult ears at two levels; all data are plotted with no SNR selection criteria imposed. The thin lines represent individual spectra while the thicker black line is a loess fit to the data to help visualize the trend. The noise floor is shown in gray. The inset to the upper panel shows SFOAEs recorded in one individual ear at 40 and 20 dB SPL, and gives an idea of the typical morphology and macrostructure of the SFOAE. The peaks, plateaus, and deep notches are intrinsic features of the SFOAE, and are associated with its fundamental generation mechanism rooted in the backscattering of wave energy off cochlear irregularity.

FIG. 2.

The upper panel shows the individual (thin gray lines) and median (thick black lines) differences between two blocks of data collected during one test session as a measure of test-retest reliability. The dashed lines around the median represent the inter-quartile 50% of difference scores from 25% to 75% and encompass differences no greater than ±1 dB. Large differences evident in some of the individual data are from measures taken at notches in the SFOAE spectra and do not represent the much greater stability of SFOAE measures taken at the peaks in SFOAE macrostructure. The inset histogram shows the distribution of difference scores with a best Gaussian fit to the peak (i.e., to the most stable part of the response). The distribution shows long tails falling outside of a normal distribution, which reflect the more outlier data measured at minima. The lower panel shows loess trend lines elucidating SFOAE level data from 15 males and 30 females. There is a significant effect of sex on SFOAE level with females having mean amplitudes 2.5 dB higher than males.

FIG. 3.

SFOAE macrostructure spacing quantified by the dimensionless ratio f/Δf (see text) so that larger numbers represent more closely spaced peaks. A loess trend line (solid) and its 95% confidence intervals (dashed) are included to guide the eye.

FIG. 4.

Group mean SFOAE I/O functions (±1 SD) were generated at eight center frequencies of a third-octave frequency band. Three exemplars (1248 Hz, 2496 Hz, and 4992 Hz) are shown here. A fit (black line) to the I/O function was used to derive estimates of the CT, defined as the stimulus level at which the SFOAE begins to grow compressively. The CT and the 95% confidence intervals are provided for each of these three examples. The dashed line shows system distortion (measured in an ear simulator), whereas the thin black line is the mean noise floor centered at the test frequency band.

FIG. 5.

SFOAE phase versus frequency functions at two probe levels for all 45 subjects with loess trend lines superimposed on each set of data. The SFOAE phase rotates through approximately 30–36 cycles across the measured frequency range. The lower level probe (20 dB SPL) produces a steeper phase slope, and thus a longer phase-gradient delay, than the 40 dB SPL probe.

FIG. 6.

Normalizing SFOAE delays: A typical SFOAE phase versus frequency function is shown in the upper panel. A delay is calculated from the phase as τ(f) = –dϕ_SFOAE/df, where ϕ_SFOAE(f) is the SFOAE phase in cycles (middle). The delay is then normalized and converted to dimensionless units by calculating the equivalent number of periods of the stimulus frequency: N_SFOAE = fτ(f). The bottom panel shows N_SFOAE from 45 young-adult ears (gray dots) at 40 dB SPL with a loess line superimposed to elucidate the trend.

FIG. 7.

(Color online) Loess trend lines fit to the normalized SFOAE delay, N_SFOAE, from 25 ears as a function of frequency with probe level (20–60 dB SPL) as the parameter. The dashed lines are the 95% confidence intervals. The individual delay values are not presented so as to more easily visualize the effects of stimulus level on N_SFOAE. There is a clear effect of probe level on SFOAE delay, with N_SFOAE generally increasing at lower stimulus levels. At the lowest probe levels, where cochlear mechanics become approximately linear, the delays are independent of level.

FIG. 8.

The top panel displays individual N_SFOAE values across frequency at 40 dB SPL. The delays are fit with two intersecting power-laws (i.e., straight lines on log-log axes) where the point of intersection is a free parameter representing the frequency at which the slope of the SFOAE delay changes (e.g., the location of the putative apical-basal, a|b, transition). The group a|b transition was estimated, as were the power-law slopes of the N_SFOAE function above and below this frequency. The bottom two panels show the group means for each of these parameters. The mean a|b seam was around 1.1 kHz for all probe levels. The low-frequency (LF) slope appears to steepen with increased stimulus level, while the high-frequency (HF) slope remains roughly invariant across level.

FIG. 9.

The upper and middle panels show SFOAE spectra and phase versus frequency functions measured with various numbers of concurrent frequency segments in one ear. The concurrent sweeps with five and six segments have starting frequencies less than one octave apart and produced noticeable deviations from our gold standard, the single-sweep SFOAE. The lower panels show the differences between the level and phase of single-sweep SFOAEs versus four-segment SFOAEs in three subjects (each line-type represents a single subject). In general, the differences between single-sweep and four-stacked segments were minimal and comparable to typical test-retest repeatability within an ear.

FIG. 10.

(Color online) The top panel shows 8 averages of the SFOAE level (each comprised of 32 sweeps) evoked with a 60 dB SPL probe and 75 dB SPL suppressor-tone in one ear. Using a conventional application of the LSF, where only the response at the probe frequency is modeled to estimate the SFOAE, incomplete cancellation of the suppressor tone in the measured residual contaminates the LSF estimate of the SFOAE; the ripples riding atop the larger macrostructure give evidence of this contamination and are expanded in two places for easier visualization. The middle panel shows eight averages of the SFOAE level after applying an LSF analysis that also includes a model of the contaminating suppressor tone. The microstructure in the SFOAE is gone and the response is highly replicable in the eight averages. The lowest panel shows estimates of the contaminating suppressor tone extracted from the measured residual.

FIG. 11.

The AR1 procedure applied in this study involved a real-time tester-directed thresholding of the time waveform recorded at the microphone; the goal was elimination of roughly 10% of the data but the threshold was adjusted throughout the test as needed. This technique was compared with a more objective offline AR method. The thick black line in this figure is the median of the SFOAE and the thin lines bracketing this value display ±1 SD, both values are dynamically updated throughout the test session. The thin gray lines are the SFOAE estimates for each individual sweep. To compare both AR methods, one should compare the white line and the dashed black line. The white line shows the complex average of the SFOAE based on the online tester-directed technique. The dashed line shows the alternative strategy, where any artifactual spike exceeding two SDs from the median is considered an artifact and triggers an additional sweep. Offline, the noisiest sweeps are eliminated (i.e., those furthest from the median SFOAE). When 10% are eliminated in this ear (dashed black line), the SFOAE estimate is indistinguishable from the tester-directed method but puts no burden on the tester. For noisier subjects, more data may be rejected offline as needed.