2D-CNN Model for Classification of Neural Activity Using Task-Based fMRI

In the recent years, several deep learning techniques have been applied to classify the cognitive states of the human brain based on the. We consider the 2D-CNN approach to classify the neural activities with the task-based fMRI. Task-based fMRI displays the neuronal activation by the blood oxygen level dependence (BOLD) response to a specific task. Spontaneous signal fluctuation with low frequency (< 0.1 Hz.) in fMRI occurs in the BOLD signal in a specific region of the human brain during the cognitive task, and voxel changes the regulation of blood flow in the brain causes the hemodynamic signal. In this paper, we have proposed a 2D-CNN model that extract the feature maps and classify the neural activity from fMRI data. The neural activation of seed to voxel connectivity in fMRI voxel are used for the classification of neural responses from the voxel. The classification performance of the proposed 2D-CNN model has been achieved from the task-evoked fMRI data with classification accuracy of 85.3%, sensitivity of 89.5%, and F1-Score of 87.2%. The experimental results shows that the proposed model effectively distinguishes the neuronal response under the task evoked stimuli.