Adaptive Radiotherapy for Head, Neck, and Thorax

Return to Projects List

Lung Segmentation

Key Investigators

• GaTech: Ivan Kolesov, Vandana Mohan, and Allen Tannenbaum
• MGH: Gregory Sharp

Objective

First, we are developing an algorithm to perform segmentation of a number of structures in the Head, Neck, and Thorax. In an effort to avoid offsetting the main advantage, accuracy, of using proton therapy, therapy plans that are based on imagery collected prior to the first session must be corrected over the course of the treatment. Once segmentation is available, the goal is to register patient scans to account for anatomical changes between visits.

Approach, Plan

• Include a shape prior model.
• Include segmented structures as prior information in the form of landmarks.
• Demonstrate an accurate segmentation on a handful of structures.
• Create a strategy for coupling segmentation and registration tasks.
• Organize/pre-process as many cases as possible.
• Create an outline for further progress after the project week.

Progress

• A pipeline for predicting segmentation of structures based on previous segmentations(later used as shape priors) exists; some synthetic experiments show its validity.
• During the project week, the speed of the algorithm was much improved.
• A part of the data pre-processing pipeline has been completed on real data sets.