Advanced Machine Leaning Prediction of Cognition Functions
Shobhika, Prashant Kumar, Sushil Chandra · 2024
Deep learning efficiently addresses difficult categorization issues in various domains, including brain state studies. End-to-end learning in convolutional neural network (CNN) based deep learning models has increased interest in computer vision. This research proposes a CNN framework to categorize the cognition function data into ‘low’ and ‘better’ states based on the cognition battery scores, namely Psychology Experiment Building Language (PEBL). To our knowledge, this is the first study to classify the meditative and non-meditative states based on the scores of PEBL tasks. Fifty-four college students were assessed using four cognition tasks before and after practicing the blend of Yoga and Rajyog meditation intervention. The proposed CNN model achieved 74% accuracy with 84% precision, 68% recall, and 75% f1-score. Thus, CNN is a viable objective measure of such data. The efficiency can further be enhanced using large-scale and multimodal data. The suggested approach may also be expanded to provide a precise categorization system for differentiating between meditators and non-meditators.