Home // HEALTHINFO 2025, The Tenth International Conference on Informatics and Assistive Technologies for Health-Care, Medical Support and Wellbeing // View article

Discrimination by Deep Learning of 1Hz Difference in Auditory Cortex using fMRI Activation Patterns

Authors:
Yoshitaka Ooyashiki
Kyoko Shibata

Keywords: fMRI; CNN; Brain decoding; Tonotopy; Region of interest.

Abstract:
Brain decoding is a technology that interprets physical and psychological states from brain activity, and it is expected to serve as a means of medical support and communication for people with disabilities. Recently, brain decoding has gained considerable attention, especially with the advent of deep learning techniques. This study builds on the concept of tonotopy in the auditory cortex and aims to develop a method to discriminate between two sounds with a 1 Hz difference, which is difficult for humans to distinguish, using brain activation images. In a previous work, the focus was on brain activation imaging acquisition methods, and research was conducted using the two main imaging designs in functional Magnetic Resonance Imaging (fMRI) experiments: event-related design and block design. The findings indicated that both designs were effective, and further improvements in accuracy are anticipated. Therefore, this report aims to further improve discrimination accuracy. To improve accuracy, this report focused on Region of Interest (ROI) expansion, hypothesizing that an increase in activation information contributes to improved accuracy of deep learning models. In this report involved the execution of experiments in which Brodmann Areas (BA) 22 was introduced as an additional ROIs, in conjunction with the existing ROIs, BA41 and BA42. The results demonstrated that expanding the ROI improved accuracy across both designs. Notably, the block design yielded an over 30% improvement, reaching 100% discrimination accuracy. The results demonstrated that ROI expansion is an effective method for enhancing accuracy.

Pages: 1 to 4

Copyright: Copyright (c) IARIA, 2025

Publication date: October 26, 2025

Published in: conference

ISSN: 2519-8491

ISBN: 978-1-68558-312-5

Location: Barcelona, Spain

Dates: from October 26, 2025 to October 30, 2025