UBC Undergraduate Research

Automated Tumor Segmentation in PET/CT Scans of Lymphoma Patients Yin, Leyi (Bellinda)

Abstract

Lymphoma is a heterogeneous disease that can manifest in over 500 lymph nodes throughout the body in addition to several lymphatic organs like the bone marrow and spleen. Assessment of disease burden, staging and prognosis is typically done by visually assessing full-body PET/CT scans. With hundreds of tumors possible, manual detection and delineation of each individual lesion is extremely time-consuming, prone to inter- and intra-observer variability, and only results in a qualitative analysis, leaving out vital quantitative metrics. Although not routinely performed in clinics, full segmentation of patient tumors can provide valuable quantitative metrics (such as metabolic tumor volume) that aid in predicting outcome and developing individualized treatment. A fully automatic pipeline for PET/CT lymphoma images for detection and delineation of the lymphoma lesions is therefore, of great importance. In recent years, with the advancement of Artificial Intelligence (AI), numerous segmentation schemes based on supervised deep learning models have been proposed that require a large number of detailed delineated cases. While detection techniques only need roughly annotated data, they can not be used to extract exact tumor boundaries. For this thesis, an automatic segmentation pipeline based on conventional segmentation methods will be implemented in MATLAB with a focus on optimizing delineation algorithms. These results will then be used to refine the output of AI-based object detection techniques, such as YOLO network. Using this routine, the work of this thesis intends to present an accurate technique for individual lesion segmentation and enhance the diagnostic workflow.

Item Citations and Data

Rights

Attribution-NonCommercial-NoDerivatives 4.0 International