BLRD Research Career Scientist Award Application

Project Summary/Abstract Dr Madabhushi has emerged as a pioneer in the development and application of novel and interpretable Artificial Intelligence (AI) algorithms for disease diagnosis, prognosis and prediction of treatment response for a variety of diseases including several cancers, cardiovascular, kidney and eye disease. Veterans, in many cases on

account of their exposure to wartime environments and particular lifestyle choices, engenders different disease phenotypes compared to the civilian population. Over the last three years he has been optimizing and tailoring AI tools to addressing problems in precision medicine for Veterans. While his primary focus has been on

diagnosis, prognosis and prediction of treatment response of lung, oropharyngeal, breast and prostate cancers for the Veteran population, he is also focused on translating and deploying these clinical decision support tools across VA stations and VISNs so that Veterans can experience precision medicine across different diseases.

Dr Madabhushi's research within the VA began in 2019 with a VA Merit award (I01BX004121) focused on AI based lung cancer screening for VA patients, specifically helping to discriminate malignant from benign nodules on routine CT scans. This work has led to development of AI driven imaging biomarkers for predicting response

to immunotherapy for lung cancer patients. More recently in a paper just published in the J of Immunotherapy for Cancer1, Dr. Madabhushi's group demonstrated the utility of radiomics on CT scans to identify clinical outcome for Stage III lung cancer patients treated with chemo-radiation therapy and immunotherapy.

Interestingly, the work showed that a subset of patients identified by his AI-based approach might be able to avoid chemo-radiation therapy and hence the associated toxicity. The study included a cohort of 15 patients from the Cleveland VA. Similarly, his team has been developing and applying AI tools both for digital pathology as

well as on radiology scans for risk stratification of oropharyngeal cancers within the VA. This work was achieved through collaboration with Vlad Sandulache at the Houston VA and with Stephen Connelly at the San Francisco VA that has resulted in a series of high impact manuscripts (J of NCI, J of Clinical Investigation, Modern

Pathology) and an NCI funded R01 (R01CA249992). In order to expand his work and footprint within the VA, he and his team have received funding support (in 2021) from the Cooperative Services Program to create a VA Hub for Computer Vision and Machine Learning in Precision Oncology (CoMPL). This new VA Hub will create computer vision and machine learning (CVML) tools

for addressing cancer diagnosis, prognosis, risk stratification and prediction of treatment response in the VA population. The objectives of CoMPL are: 1) focus on building the computational infrastructure and tools to allow for expanding the scope and access to CVML resources within the VA, and building a community to enable VA

researchers to take advantage of these tools to develop their own CVML applications; and 2) to develop new companion diagnostic tools for risk assessment, predicting response and need for more or less aggressive therapy in prostate and lung cancer. An initial demonstration project of CoMPL will focus on application of AI

tools with CT scans and digital pathology images to identify benefit of adjuvant chemotherapy in early-stage Veteran lung cancer patients. Dr. Madabhushi's is also leading a new prostate cancer collaborative involving urologists, radiologists and oncologists from multiple different VA stations and VISNs to develop the use of AI

with multimodal imaging (MRI and digital pathology) along with genomics for more accurate risk stratification of Veterans with high-risk prostate cancer. The CoMPL team is partnering with the National Artificial Intelligence Institute (NAII), Lung Cancer Precision Oncology (LPOP) and Precision Oncology Program for Cancer of the

Prostate (POPCaP) centers to enable dissemination of the decision support tools and the deeply annotated digital pathology and radiology scans that result from CoMPL's activities.