STTR Phase I: Automated object contouring methods and software for head and neck radiotherapy planning

Period of Performance: 01/15/2016 - 12/31/2016

$225K

Phase 1 STTR

Recipient Firm

Quantitative Radiology Solutions LLC
PHILADELPHIA, PA 19104
Firm POC, Principal Investigator

Research Institution

University of Pennsylvania
3330 Walnut Street, Levine hal
Philadelphia, PA 19104
Institution POC

Abstract

The broader impact/commercial potential of this Small Business Technology Transfer (STTR) Phase I project is in the field of radiation therapy (RT) for head & neck cancer patients. RT planning involves designing a radiation treatment regimen so that tumors are irradiated to definitive doses while minimizing irradiation to normal structures. For devising an optimal RT plan, target tumors and critical anatomic structures need to be accurately contoured on medical images. In current clinical practice, organ contour delineation is performed mostly manually due to lack of automated contouring software. This makes RT planning error prone, hampers throughput, and does not allow re-contouring to handle changes taking place during RT. Such changes can cause under dosing to tumor and overdosing to normal surrounding organs. In 2015, 1,658,370 new cancer cases are estimated to occur in the US, where nearly two-thirds will have RT. Given that there are over 2,100 RT centers in the US, there is a strong commercial opportunity for producing an auto-contouring software system. Expected clinical outcomes are significantly improved speed, throughput, and accuracy of contouring compared to current clinical practice, and improved patient outcomes and cost-savings. This Small Business Technology Transfer (STTR) Phase I project addresses a technical hurdle related to auto-contouring in RT planning for cancer patients. Current technical challenges for auto-contouring occur since available contouring methods have been developed mostly for a specific object on images of a particular modality. This project will overcome these hurdles through a novel automatic anatomy recognition methodology which will employ anatomy models derived from patient populations by including all major objects in a body region. The models will codify the rich object anatomic relationship, and will exploit this information to automatically locate and contour objects in any given patient image. The project will have two aims. Aim 1 involves the development of the method and prototype software for contouring major head & neck organs on CT and PET/CT images. Models will be built from already existing image and contour data of 200 cancer patients. Aim 1 outcome will be prototype software technically validated to be accurate within 1 pixel boundary distance compared to ground truth and requiring 3 minutes or less per study. Aim 2 will be a preliminary clinical assessment of the software in RT planning in patients with head & neck malignancies.