APPLICATION OF FEATURE CORRELATION ON MEDICAL BIG DATA ANALYSIS USING DEEP LEARNING

摘要

This research describes a fully intelligent and automated brain tumor segmentation system depending on Deep Convolutional Neural Networks with feature correlation. Hierarchical CNN model motivated the system design, which was tweaked to improve brain tumor detection performance. Both the BraTS' 2018 training and testing datasets were used to implement and refine the proposed method. Additionally, we have compared the two models i.e; with feature correlation learning model and without feature correlation learning model on the basis of accuracy and model loss and it revealed that with feature correlation learning model achieved the accuracy of 99 with minimum loss value of 2. This shows that transfer-learning model outperforms without feature correlation learning model.