Rapid neonatal hearing screening using a modified maximum length sequences automated auditory brainstem response
Ahmad Aidil Arafat Dzulkarnain · The University of Queensland · 2008
The automated auditory brainstem response (AABR) test is an important tool for detecting hearing impairment in newborns in Universal Newborn Hearing Screening (UNHS) programs, but it is often hindered by unacceptably long test times. Currently, a typical AABR test takes approximately three to seven minutes per newborn to complete. These test times contribute significantly to the cost of UNHS when millions of newborns need to be tested each year in UNHS programs around the world. The aim of this thesis was to determine if high stimulus repetition rates combined with automated signal detection algorithms could reduce AABR test times in newborns who pass the AABR assessment in the UNHS setting. To resolve this aim, the thesis study proceeded in three phases. Phase 1 describes the background to the study’s aims, including UNHS, the ABR and AABR, high stimulus repetition rates, maximum length sequences (MLS), and methods of automated signal detection. The conclusion was that a reduction in the AABR test times would be useful for UNHS, and that this might be achieved by using high stimulus repetition rates and automated signal detection. Phase 2 describes the construction, calibration and testing of a dedicated PC based auditory evoked potential (AEP) system that could be controlled by the researcher using Matrix Laboratory (MATLAB) software. The results of this phase were that this system could record ABR waveforms from adult and newborn subjects using both conventional and MLS click stimuli. It was therefore concluded that the newly constructed PC-based AEP system could be used in Phase 3 of the thesis study. Phase 3 describes the use of the newly constructed PC-based AEP system to record AABR waveforms from 78 newborn subjects who had passed a UNHS assessment at a large birthing hospital in Brisbane, Australia. These AABR waveforms were recorded using conventional click stimuli at 33 clicks-per-second (cps) and 90 cps, and Maximum Length Sequence (MLS) click stimuli at 180, 250, 500 and 833 cps (maximum stimulus rates) with the MLS AABR reconstructions conducted using both a standard bipolar MLS reconstruction and a novel unipolar MLS reconstruction. Each resulting response was analysed for the presence or absence of an AABR waveform, and, if present, for the time to detection of the ABR waveform using both Fsp and cross-correlation analyses at confidence levels of 99%, 95% and 90%. Any significant differences in these times to detection were determined using ANOVA and Tukey’s Honest Significant Difference analyses. The results showed that none of the combinations of stimulus repetition rates, MLS reconstruction techniques, Fsp or cross-correlation analyses, or significance levels, showed 100% AABR waveform detection in 100% of the newborn subjects. The pass rates ranged from as low as 3.85% for the AABR waveforms obtained using MLS stimuli at a maximum stimulus rate of 833 cps with a bipolar or a unipolar reconstruction and detected using cross-correlation analysis at a 99% confidence level, to as high as 87.12% for AABR waveforms obtained using MLS stimuli at a maximum stimulus rate of 500 cps with a unipolar reconstruction and detected using cross-correlation analysis at a 90% confidence level. Of those AABR waveforms deemed to be present, the times to detection ranged from as low as 6.28±6.76 s for AABR waveforms obtained using MLS stimuli at a maximum stimulus rate of 833 cps with a unipolar reconstruction and detected using Fsp analysis at a 99% confidence level, to as high as 40.03±18.53 s for AABR waveforms obtained using conventional stimuli at 33 cps and detected using cross-correlation analysis at a 99% confidence level. The best balance of high pass rates with low times to detection was considered to have occurred for AABR waveforms obtained using MLS stimuli at a maximum stimulus rate of 500 cps with a bipolar or unipolar reconstruction and detected using cross-correlation at a 90% confidence level. This combination gave pass rates of 85.9% and 87.12% respectively and times to detection of 14.13±8.50 s and 14.27±8.27 s respectively. This result differed from previous results showing that MLS stimuli at the maximum stimulus rates of 200 to 300 cps were preferred for adult subjects. This suggests the auditory system of the newborn subjects was more linear than that of the adult subjects, thus allowing it to benefit more from these higher stimulus rates. No significant benefit was gained by using the novel unipolar MLS reconstruction versus the standard bipolar MLS reconstruction. This thesis concluded that high stimulus repetition rates combined with automated signal detection algorithms can reduce the AABR test times in the UNHS setting, but further research is needed to improve the rates of the AABR waveform detection at high levels of confidence. In this regard, better methods of MLS reconstruction (for example, non-linear reconstruction), artifact rejection and automated signal detection are needed if these improvements are to be achieved. Any efforts in this regard should concentrate on the MLS ABR at 500 cps, because this was the best performing stimulus in the newborn subjects assessed in this thesis study.