Executive Summary : | Early diagnosis of a brain tumor can improve the patient's chances of recovery after therapy. Medical imaging technologies have advanced significantly in the previous decade, and they are now playing an increasingly important role in diagnosis and treatment. AI-powered detection tools, for segmentation, grading and lifetime prediction play a very significant role. Early detection of brain tumours is critical for improving patient survival and treatment success. Manually evaluating magnetic resonance imaging (MRI) images is a tough undertaking. As a result, more accurate digital approaches for tumour diagnostics are required. However, determining their structure, volume, borders, tumour detection, size, segmentation, and classification remains a difficult task. High variability and inherent MRI data properties, such as variability in tumour sizes or shapes, tumour identification, area computation, segmentation, classification, and discovering ambiguity in the segmented region, make Brain Tumor a difficult task. Choosing the correct deep learning tools is particularly difficult because it necessitates an understanding of numerous parameters, training methods, and topologies. The hybridization idea is very essential to deal with uncertainty. In this proposed work, novel pre-processing techniques, segmentation architecture(Core Tumor(CT), Enhancing Tumor(ET), Edema (ED), Non-Enhancing Tumor(NET)) will be developed to mark the different regions tumor. Further image-based grading using the 3d CNN model and feature-based grading using Recurrent neural and the result will be fused in a Bayesian belief network format. Meta-analysis will be explored to improve the accuracy of the decision. The meta-analysis, also known as evidence synthesis, is a method for synthesising and integrating findings from related research to draw a generalizable conclusion regarding the efficacy of a clinical strategy from a wider pool of data. The odds ratios, risk ratios, and risk differences are among the most widely used clinical effects in meta-analysis. The use of meta-analysis for summarising clinical evidence has increased dramatically during the last few decades. Meta-analysis has now become a hallmark of evidence-based medicine as a result of its widespread application. Further, lifetime prediction using reinforcement learning techniques will be used for continuous learning and updating. The obtained accuracy for lifetime prediction in the literature is less. The accuracy rate will be improved and a more reliable tool will be developed using a reinforcement learning framework. |
Co-PI: | Dr. Ramkumar Kannan, Sastra University, Tamil Nadu-613401, Dr. Kanagasabai Adalarasu, Sastra University, Tamil Nadu-613401, Dr. R Raghavendran, Madras Medical College, Chennai, Tamil Nadu-600003 |