Overview

Cancer is one of the leading causes of death worldwide, causing nearly 10 million deaths in 2020, or almost one in six. The heterogeneity in cancer physiology between patients and the cytotoxic nature of oncology therapies requires the care providers to optimize and personalize the drug choices and dosimetry to maximize the disease-free survival rate. Molecular-level characterization of cancerous tumors heavily improves the treatment strategies by targeting genetic mutations through individualized therapy. Theranostics is a novel therapy regime where pairs of radiopharmaceuticals are used for diagnostic and therapeutic purposes. The success of this therapy relies on identifying the optimal care plan for each patient-level factor, tumor burden, Organs At Risk (OAR) radiation tolerance, and complex biological clearance and uptake of the therapeutics. To address this problem, our research aims to advance the current state of the art in three separate but interdependent areas within the therapy workflow: First, we will develop a data-driven Physiology-Based PharmacoKinetic (PBPK) model to estimate patient-specific parameters from the PBPK model, individual patient biodistribution data, and features extracted from diagnostic imaging. Second, we will develop a deep learning model for the estimation of absorbed dose based on the PBPK model. Time-Integrated Activity (TIA) from the PBPK model will predict the absorbed dose at both OAR and tumors. Third, we will investigate the application of Reinforcement Learning (RL) to the optimization of treatment procedures for the partially observable settings of this problem while also optimizing for multiple objectives, including treatment efficacy and reduced impact on OAR.

Team

Binesh Kumar

Advisor: Dr. Vahid Behzadan

Affiliates: N/A

GitHub: N/A

Publications: N/A

Sponsor: N/A