Optimizing swept-tone protocols for recording distortion-product otoacoustic emissions in adults and newborns

Abstract

Distortion-product otoacoustic emissions (DPOAEs), which are routinely used in the audiology clinic and research laboratory, are conventionally recorded with discrete tones presented sequentially across frequency. However, a more efficient technique sweeps tones smoothly across frequency and applies a least-squares-fitting (LSF) procedure to compute estimates of otoacoustic emission phase and amplitude. In this study, the optimal parameters (i.e., sweep rate and duration of the LSF analysis window) required to record and analyze swept-tone DPOAEs were tested and defined in 15 adults and 10 newborns. Results indicate that optimal recording of swept-tone DPOAEs requires use of an appropriate analysis bandwidth, defined as the range of frequencies included in each least squares fit model. To achieve this, the rate at which the tones are swept and the length of the LSF analysis window must be carefully considered and changed in concert. Additionally, the optimal analysis bandwidth must be adjusted to accommodate frequency-dependent latency shifts in the reflection-component of the DPOAE. Parametric guidelines established here are equally applicable to adults and newborns. However, elevated noise during newborn swept-tone DPOAE recordings warrants protocol adaptations to improve signal-to-noise ratio and response quality.

FIG. 1.

(A) and (B) DPOAE recorded with swept tones and analyzed in one adult ear (sweep rate: 0.5 octave/s, analysis-window duration: 125 ms). The thin gray lines represent DPOAE level and phase estimates from each of 24 individual sweeps. The thicker black line is the complex average derived from the individual sweep-based estimates. The noise floor is shown as dotted light gray. (C) and (D) DPOAE from one newborn ear recorded and analyzed in the same way as described for the adult ear, however, with more sweeps (n = 32). The thin gray lines in all panels depict the variance of repeated DPOAE estimates within an ear and provide a measure of intra-subject response stability.

FIG. 2.

The standard error of mean (SEM) for adult and newborn estimates of DPOAE level. This intra-subject SEM was calculated for all subjects from repeated sweep-based estimates [see Figs. 1(A) and 1(C)], first in pressure and then converted to dB SPL. This procedure allows for a measure of within-ear stability and provides a metric of response quality for each subject. The group SEM was determined by fitting a loess trend line to the individual functions; it approximates group-noise floor measurements, which are shown as dotted lines. In adults, the group noise floor is within a few dB of the calculated intra-subject SEM. In newborns, within-ear variability is greater than in adults and the SEM is notably higher than the group noise floor suggesting other sources of variance.

FIG. 3.

Intra-subject SEM measures provide a metric of response quality or test-retest reliability within an ear and can be converted into 95% confidence intervals (CI) around the complex average. As an example, sweep-based CIs are shown here for DPOAEs recorded in one newborn ear.

FIG. 4.

(Color online) DPOAE level and phase recorded with a fixed sweep rate of 0.5 octave/s and analyzed with five analysis-window durations in one adult and one newborn subject. The longer windows smooth the DPOAE excessively and do not preserve fine structure. The shortest window produces additional oscillations riding atop the larger fine structure. Both 62 and 125 ms preserve DPOAE fine structure well and appear to be optimal analysis windows for this sweep rate. The analysis bandwidth associated with each window condition is shown by the length of the horizontal bars plotted near the frequency axis; the longer the window, the broader the bandwidth (and the smoother the DPOAE).

FIG. 5.

Time-domain analyses of DPOAE in one adult and one newborn ear for three frequency segments. The longer analysis windows smooth the total DPOAE by eliminating contributions from the reflection component as shown in these exemplars. The large dominant lobe around 0 ms is the distortion component of the total DPOAE and is equally present for analyses with both a short (125 ms) and long (500 ms) window. The smaller lobe noted at ∼5–10 ms is the reflection component of the DPOAE, and its energy is reduced when analyzed with a 500 ms window. The longer window produces spectral smoothing and reduces reflection energy in the DPOAE.

FIG. 6.

DPOAE level (top panels) and phase (bottom panels) in three half-octave bands for one adult subject. DPOAEs were collected and analyzed using traditional discrete-tone presentation with FFT analysis (small gray dots) and swept-tones with LSF-based analysis and the two optimal window durations: 62 and 125 ms (or 31 and 62 ms for the higher-frequency segment). The swept-tone analysis approximated discrete-tone DPOAE fine structure within a few dB, and the difference in accuracy of the match among analysis-window conditions was statistically insignificant.

FIG. 7.

The top panel shows the effect of analysis-window duration on DPOAE noise floor. The bottom panel shows the effect of analysis-window duration on intra-subject SEM measures for the adult and neonatal age groups. The longer the analysis window, the lower the noise floor, the better the SNR (data not shown here), and the more stable and repeatable the response within an ear. Error bars represent ±1 standard deviation.

FIG. 8.

DPOAE level in one adult ear to illustrate the effect of sweep rate on DPOAE fine structure. (The analysis bandwidth was fixed at 0.06 octaves for each rate and optimal analysis-window durations were applied.) Fine structure peaks often shift as sweep rate is increased from 0.125 to 1 octave/s. The inset magnifies this change for one of these peaks. The average peak shift was 3–7 Hz for adults and 6–12 Hz for newborns, suggesting an effect of sweeping rate on OAE phase.

FIG. 9.

(Color online) The inset displays data from two previous studies showing reflection-component delays (Abdala et al., 2014) and stimulus-frequency OAE delays (Shera et al., 2010) expressed in periods. Both confirm longer delays with increasing frequency. To accommodate the shifting latency of the reflection-component, analysis-window duration (and bandwidth) must change with increasing frequency. (A)—(D) show DPOAE level and phase recorded and analyzed with a fixed sweep rate of 0.5 octave/s and five analysis-window durations in one adult (A) and (B) and one newborn (C) and (D) for frequencies from 4 to 8 kHz. The trends are similar to those shown in Fig. 4 except that shorter analysis windows and narrower bandwidths are required for DPOAEs at higher frequencies. Analysis windows of 31 ms and 62 ms best preserve DPOAE fine structure. Following the convention of using the longest analysis-window duration that captures the total DPOAE, among these options, the optimal analysis window for the 4–8 kHz data is 62 ms (which corresponds to a 0.03 octave bandwidth).

FIG. 10.

Although this study focuses on the optimal parameters for the collection of the total DPOAE (including both distortion and reflection components), a rapid way to isolate the distortion component from the reflection component of the response is to use a long window, 500 ms, with a too-fast sweeping rate, 0.5 octave/s. This effectively isolates the shorter-latency distortion energy from the later reflection energy. The main panel displays DPOAE fine structure from one adult ear (dotted line) and the distortion component of the DPOAE extracted using an inverse FFT (gray line) versus the above-described LSF-based method (black line). The two distortion-component measures are nearly indistinguishable. This is further elucidated for group data by the inset, which displays individual differences in dB for all adult subjects as thin gray lines and a loess trend line superimposed in black. The average overall difference between IFFT and LSF-based estimates of the distortion component was 0.4 dB.

FIG. 11.

The inset in this figure illustrates the desired function between frequency and both analysis-window duration (Δt) and associated analysis bandwidth (BW) to optimize DPOAE measurements and include both components. The dashed horizontal lines represent the two fixed analysis-window durations determined to be optimal for DPOAEs above/below 4 kHz, i.e., 62 and 125 ms. This function was derived from DPOAE reflection-component delays reported in Abdala et al. (2014). The main panel of this figure displays DPOAE level analyzed with smoothly shifting analysis windows which follow the inset function (black) compared to DPOAE level analyzed using only two fixed analysis-windows (gray).

FIG. 12.

The effect of two strategies on noise estimates in 10 neonates (each ear denoted by a separate symbol). “Noise reduction” is the improvement in noise floor re: baseline (analysis window of either 31 or 62 ms depending on frequency, and sweep rate of 0.5 octave/s) The larger the dB value, the more effective the noise reduction. Strategy A shows that the doubling of analysis windows produces approximately 2 dB reductions in noise floor (as expected from data shown in Fig. 7) for all newborns across frequency. Strategy B shows data from the same newborns after applying combined modifications: (1) slowing down the sweep rate to 0.125 octave/s; (2) lengthening the analysis window to 500 ms; and (3) doubling the number of sweeps contributing to each average. These combined strategies produce additional noise floor reductions, some as great as 14 dB.