Deep Learning-Based Approaches for Machine Interface Analysis Using MRI Images

Abstract

The tumor is a lethal illness that initiates due to unregulated growing of cells in the organs of the body like the brain, iris, spleen, lungs, etc. It is crucial to make an early diagnosis. There are numerous medical imaging methods, including CT, PET, Ultrasound, and MRI and so on, but MRI is frequently approached modality for its less ionization and less radiation. Recently, DL techniques are more popular in medical imaging technology. The important contribution of the article is to compare the effectiveness of the deep learning types/techniques for detecting any tissue from the T1-weighted (T1w) MRI abnormal brain images. In this article, the most used DL methods like CNN, RNN, and DBN are used, and analyzed the performance of each method in terms of DSC (dice score coefficient), PPV (positive predictive value), and sensitivity by using the BraTS 2020 dataset. The results of this segmented image have obtained the scores of 0.89, PPV of 0.87, and sensitivity of 0.90 for the CNN method. The C.N.N-based method is more effective for the brain tumor detection than RNN and DBN techniques.