Executive Summary : | The project proposes the design and development of novel Machine learning and Deep learning-based algorithms for development of an automated real-time decision support system for the detection and grading of OSCC from H&E Histopathology microscopic images. In India, oral cancer is a critical medical issue since it affects a larger portion of the population. Annually in India, around 77,000 new cases and 52,000 deaths are recorded. Contrary to the west, in India, 70% of cases are reported in the advanced stages, leaving a survival rate of around 20%. Oral Squamous Cell Carcinoma (OSCC) is the most prevalent type of oral cancer and accounts for 84-97 % of worldwide cases. Diagnostic methods for oral cancer include physical examination, histopathological examination, and imaging techniques such as MRI. Histopathology analysis is the gold standard of OSCC analysis, in which pathologists examine tissue sections or biopsy samples under a microscope for diagnosis and prognosis. If diagnosed with OSSC the pathologist grades the tissue section. Grading refers to the degree of differentiation between macroscopic and microscopic of the tumor, and the World Health Organization endorses a simple grading system based on Broaders' criteria. Gradings are as follows: Well-differentiated (up to 25% anaplastic), Moderately Differentiated (25 - 70% anaplastic), and Poorly Differentiated (more than 75% anaplastic). By automating the process, the proposed project will contribute to mitigating the drawbacks of manual histopathology evaluation by reducing Observer bias, reducing time consumption, minimizing the high false-positive rate, and assisting in better prognosis, especially in developing countries, where the patient is to doctor ratio is poor. The proposed project will involve the design and generation of a multisource and exclusive Dataset of OSCC microscopic Histopathology, A study of architecture level histology features like keratinization, presence of keratin pearls, break-in basement membrane, etc., Development of Image Pre-processing Techniques such as luminous standardization and Stain normalization, Design and development of segmentation technique, which will be instrumental in identifying the histopathology features such as break-in basement membrane pattern, identifying keratin pearls which will help in counting the keratin pearls and also segment out the mitosis in OSCC histopathology images followed by design and development of Machine learning and Deep learning classifiers for detection (binary classification) and grading(multiclass classification) of the OSSC histopathology images |