NeurIPS 2019 Expo Demo

Dec. 8, 2019

Expo 2019 Schedule »

Deep Learning-Based End-to-end Automatic Contouring and Automated Radiation Therapy Treatment Planning System

Sponsor: BioMind

Abstract:

For patients suffering from cancer, surgery may not be a feasible solution and radiation therapy is a common alternative treatment method. Radiation therapy is the treatment of cancers via the use of targeted high-energy radiation, which may be delivered externally or internally. Performing such treatment involves creating a treatment plan for implementing the radiation dose onto the patient, which requires radiologists to manually delineate the boundaries of patients’ tumors and organs-at-risks, and iteratively adjust the constraints and weights of the dose objective functions relating to the delineated tumors and organs-at-risks. However, the process is mostly manual, making it time-consuming and inconsistent. This is undesirable considering that a poorly designed treatment plan could adversely affect the function of healthy organs if the organs are overdosed, or unsuccessfully remove tumors if the tumors are underdosed.

Therefore, this demonstration shows how deep learning can be applied in the field of radiation therapy, specifically the creation of a radiation therapy treatment plan from only patient medical images and prescription information from healthcare professionals. Attendees will be able to interact with the system through an easy-to-use interface, allowing them to witness the treatment planning process and observing and interacting with the automated contouring, treatment plan and patient radiation dose obtained via the treatment planning process.

The purpose of the demonstration is to show attendees that deep learning is ready to be deployed for radiation therapy, creating a positive difference by saving lives and freeing up time for radiologists to enhance their ability to provide healthcare for patients.