An Open-Source Deep Learning-Based Toolbox for Automated Auditory Brainstem Response Analyses (ABRA)

Area of Science:

• Neuroscience
• Bioengineering
• Computational Biology

Background:

• Hearing loss is a global health issue linked to cognitive decline.
• Auditory brainstem responses (ABRs) are crucial for assessing auditory pathway function.
• Manual ABR analysis is subjective and lacks reproducibility.

Purpose of the Study:

• To develop an automated tool for analyzing ABR waveforms.
• To improve the speed, accuracy, and reproducibility of ABR analysis.
• To facilitate interdisciplinary research in hearing science.

Main Methods:

• Developed Auditory Brainstem Response Analyzer (ABRA), an open-source GUI tool.
• Utilized deep learning, specifically convolutional neural networks, for waveform analysis.
• Trained models on diverse ABR datasets from multiple experimental settings.

Main Results:

• ABRA achieved rapid and unbiased extraction of ABR metrics (amplitude, latency, thresholds).
• Deep learning models demonstrated performance comparable to expert human analysis.
• Analysis time was significantly reduced, and reproducibility was enhanced across datasets.

Conclusions:

• ABRA offers a standardized, automated solution for ABR analysis.
• The tool enhances research reproducibility and efficiency in hearing science.
• ABRA bridges hearing research, neuroscience, and computational techniques for broader insights.