Schedule of Events | Symposia

Development of a Diagnostic Model for Mild Cognitive Impairment Using Brain Functional Connectivity

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Chisho Takeoka¹ (takeoka.chisho904@mail.kyutech.jp), Tetsushi Yada¹, Toshimasa Yamazaki², Yoshiyuki Kuroiwa³, Kimihro Fujino³, Toshiaki Hirai³, Hidehiro Mizusawa⁴; ¹Graduate School of Computer Science and Systems Engineering, Kyushu Institute of Technology, ²Office for Career Support, School of Computer Science and Systems Engineering, Kyushu Institute of Technology, ³University Hospital, Mizonokuchi Teikyo University School of Medicine, ⁴National Center of Neurology and Psychiatry

Proper diagnosis of mild cognitive impairment (MCI) is important in the early detection of dementia. In this study, we tested the possibility of discriminating between electroencephalography (EEG) data of MCI patients and healthy elderlies by machine learning. 16 regions of EEG data measured for 7 MCI patients and 10 healthy elderlies were separated into 6 frequency bands each, and the functional connectivity between two brain regions were calculated with Synchronization Likelihood (SL) in each frequency band. A discriminator based on 16C2 SL values was constructed for each frequency band, and the discrimination rates between MCI and healthy elderlies were calculated. Leave-one-out, leave-two-out, and leave-three-out cross-validations were performed to evaluate the discriminator, and the SL values used for cross-validation were selected in the range of 1 to 120 using Recursive Feature Elimination (RFE). The results showed that the discrimination rates with random forests using 12 SL values in the lower alpha band were 88%, 75%, and 72% for leave-one-out, leave-two-out, and leave-three-out cross-validation, respectively. The SL values that were frequently selected throughout the series of RFEs represented the functional connectivity of the frontal and left temporal lobes, indicating the robustness of the discriminator with these SL values. These indicate that it is possible to identify MCI using EEG data and that functional connectivity between the frontal and left temporal lobes is associated with cognitive dysfunction in MCI. Details will be presented in a poster presentation.

Topic Area: METHODS: Other