Title: Optimization of patient-specific neoadjuvant therapy regimens for breast cancer via image-guided fluid dynamics modeling

Abstract: Systematically delivered drugs, including chemotherapy, hormone therapy and targeted immunotherapy, play important roles in adjuvant and neoadjuvant treatments of breast cancer. Usually, physicians will make the treatment plan based on multiple considerations, like types of breast cancer, stages and any special situations, which is a quite complex decision-making process, which, obviously, has a great importance to be personalized. Specifically, the decision of drug injection protocol, involving the regimen and dose of drugs being given, as well as combination of different drugs, is an essential part in treatment planning.

Personalization of drug injection protocol for breast cancer treatment remains an actively investigated topic. Based on instructions in current clinical practice, chemotherapy is usually given in periodic cycles of 3 – 4 weeks, with combination of 2 or 3 different drugs. Within
each cycle, there are multiple options of delivery profile, for example, giving the drug with a single large pulse at once, or with multiple small injections every a few days. However, these are mostly general recommendations, and it is hard to determine which way will work the best for one specific patient. Moreover, the dose is also challenging to optimize with general recommendations or
experiences.

Mathematical and computational oncology could significantly contribute to these open questions in personalization of treatment planning. In this project, we seek to develop an image-guided modeling system patient-specific optimization of drug injection protocol for breast cancer treatment. Specifically, we establish an 1D-3D coupled computational fluid dynamic model, and take pre-treatment MRI as inputs to estimate patient-specific hemodynamics and pharmacokinetics. Then, varying combinations of practical drug injection profiles and doses, we build an objective function to maximize treatment efficacy within tumor and minimize toxic side effect outside the tumor. This modeling system can efficiently solve the personalization of drug injection protocol as an optimal control problem. The results indicate that different patients have different optimal injection protocol and react differently to the variation of injection protocols.