Predictix- Deep Learning Framework for Prospective Radio Genomic Glioma Tumors' Analysis in MRIs

Webinar on 2^nd International Congress on AI and Machine Learning

February 15, 2022 | Webinar

Anant Bhatt

Charotar University of Science and Technology, India

ScientificTracks Abstracts: RRJET

Abstract

Gliomas frequently exhibit a discernible propensity for local recurrences in resection margins, most frequently in peritumoral oedema (PTE), which stipulates an exquisite distinction from "Vasogenic Edema" (VE). Hence, the "True Extent" of tumors in PTE is a therapeutic challenge. WHO classification lists distinct genomic subtypes based on extensive methylome studies, with Magnetic Resonance Imaging (MRI) as the mainstay of radiological diagnosis. We propose an automated Artificial Intelligent Glioma analysis Framework incorporating stacked implementation named - "Predictix". The erudite Deep Learning Pipeline quantifies tumor imaging physiognomies using a deep learning stack of algorithms infused with segmentation. 2-D and 3-D characteristics from preoperative imaging superimposed with CNN layer with tranfer learning empowers precise imaging MRI classification. Genomic clusters incorporating patients' IDH mutation, 1p/19q codeletion, DNA Methylation, Gene Expression, DNA copy number, and microRNA expression are probed in the subsequent layers to examine the correlation between the chromatic imaging exposures and genomic clusters. A comprehensive pipeline maps the findings with Fisher's exact test using Bonferroni correlation, determining co-relative mapping at the end. 'Predictix' - a novel infused AI stack enables the delimitation of PTE margins, with extensive insinuations in outlining preoperative planning, which is invaluable in the motor cortex, in conjunction with newer neurosurgical techniques, i.e., ALA or neuronavigation. We envisage substantial benefits of predictive in defining reduced clinical target volume (CTV) for adjuvant radiation therapy by reducing therapeutic volume and thus long-term morbidity. It permits faster, further effective diagnosis and classification of tumors and refines itself to classify the genomic subtypes. The comprehensive pipeline, bolstered by factual patient records, is envisaged as a quantitative "realtime" feedback mechanism.

Biography

Anant has completed his Mtech from IIT Kharagpur and pursuing his PhD from Charusat University. He is the director at Governement of India, and has been founder of the Artificial Intelligence R&D centre. He has published numerous research and innovation for which he has been awarded by the Prime Minister of India twice. He has been reviewer of many renowned international journals. His unflinching, diverse career includes expertise in Electronic warfare, Satellite Image analytics and operational communications.

• PDF
• HTML