Barnett Nelson V, Hoyland Alec, Chari Divya A, Parrell Benjamin, Lammert Adam C
Department of Biomedical Engineering (BME)Worcester Polytechnic Institute (WPI) Worcester MA 01609 USA.
Clarifai, Inc. Wilmington DE 19808 USA.
IEEE Open J Eng Med Biol. 2024 Jul 18;5:589-592. doi: 10.1109/OJEMB.2024.3427318. eCollection 2024.
We validate a recent reverse correlation approach to tinnitus characterization by applying it to individuals with clinically-diagnosed tinnitus. Two tinnitus patients assessed the subjective similarity of their non-tonal tinnitus percepts and random auditory stimuli. Regression of the responses onto the stimuli yielded reconstructions which were evaluated qualitatively by playing back resynthesized waveforms to the subjects and quantitatively by response prediction analysis. Subject 1 preferred their resynthesis to white noise; subject 2 did not. Response prediction balanced accuracies were significantly higher than chance across subjects: subject 1: 0.5963, subject 2: 0.6922. Reverse correlation can provide the foundation for reconstructing accurate representations of complex, non-tonal tinnitus in clinically diagnosed subjects. Further refinements may yield highly similar waveforms to individualized tinnitus percepts.
我们通过将一种最近的反向关联方法应用于临床诊断为耳鸣的个体，来验证该方法在耳鸣特征描述方面的有效性。两名耳鸣患者评估了他们非音调性耳鸣感知与随机听觉刺激之间的主观相似性。将反应回归到刺激上产生了重建结果，通过向受试者回放重新合成的波形进行定性评估，并通过反应预测分析进行定量评估。受试者1更喜欢他们的重新合成声音而不是白噪声；受试者2则不然。跨受试者的反应预测平衡准确率显著高于随机水平：受试者1为0.5963，受试者2为0.6922。反向关联可为在临床诊断的受试者中重建复杂的非音调性耳鸣的准确表征提供基础。进一步的改进可能会产生与个体化耳鸣感知高度相似的波形。
