Abstracto

Proposal of classification methods for automated analysis combining data-driven and knowledge-driven approaches in fNIRS study

Obata A, Fukui D, Egi M, Sutoko S, Atsumori H, Funane T, Nishimura A and Kiguchi M

We propose classification methods combining data-driven and knowledge-driven approaches for automated multivariate analyses in fNIRS study. In the data-driven approach, we did training of brain activity features during a task including those suffering from daily depressive mood using regularized regression, and we obtained several prediction models with high accuracy. In the knowledge-driven approach, we confirmed previously related research findings and classified the availability of each brain activity feature (available or new) in the prediction models. Finally, models were validated, and the results showed that the accuracy using the available features and one new feature was higher than the accuracies using the available features only. A new parameter regarding depressed mood was found using the proposed approach.