Executive Summary : | In the era of precision medicine, researchers are utilizing technology to provide treatment outcomes by considering diverse data sources, including image features, clinical features, and automatically learned features from deep neural networks. Information fusion techniques have been developed to solve predictive problems, but these data representations contain distinct and interrelated information. This proposal aims to integrate this diverse source of data by handling conjunctions to model interdependency among different features. The research focuses on developing new tools for modeling and aggregating interrelated features, designing regression approaches with the adoption of interrelated features, and fusion of different prediction algorithms using MCDM techniques. The research proposes algorithms for combining multiple features by identifying independent and dependent features, and investigates a hierarchical predictive scheme to link multiple features and prediction models more logically and effectively. The goal is to build a model using the developed hierarchical predictive technique to predict dose distribution prior to initiation of treatment planning in cancer radiation therapy. Prostate cancer is one of the top ten leading cancers in India, and radiotherapy is established as an effective modality with control rates that are at par with surgery. However, there is a disparity in the definition of an optimal treatment plan leading to interpersonal variation. The proposed hierarchical predictive model will allow clinicians and physicists to predict radiation dose data prior to treatment planning, leading to the development of patient-specific dosimetry planning that doctors may alter according to their clinical judgement. |
Co-PI: | Dr. Indranil Mallick, Tata Medical Center,Kolkata, West Bengal-700160, Prof. Debjani Chakraborty, Indian Institute Of Technology (IIT) Kharagpur, West Bengal-721302, Dr. Arnab Sarkar, Indian Institute Of Technology (BHU), Varanasi, Uttar Pradesh-221005 |