Whole body soft-tissue lesion matching
One general aim of our research group is developing tools to help us identify factors that are predictive of patient response to cancer treatments. Previous collaborators have developed an articulated bone registration software to facilitate quantitative analysis of bone metastases.
My work in the research group is to expand the existing software, so that we can dispose of a similar framework to quantitatively analyze progression of soft-tissue cancer metastases, for example melanoma. The development is based on datasets from patients of different cancers that received whole body medical-imaging PET/CT exams. To assess the disease response to treatment, each patient received many scans, at different time-points during cancer treatment.
My goal is to develop an automatic computerized image analysis framework that extracts relevant quantitative information from these scans. This framework includes: image segmentation of the patients’ body regions, segmentation of different tissues within the body regions, detection and segmentation of lesions and lesion matching between different time points for treatment response assessment. The latter task includes techniques such as soft-tissue deformable registration and deep neural network classification.
Besides helping us in the cancer response characterization research, such tools can be used in clinics to allow for personalized medicine. Automated treatment response analysis allows physicians to assess individual patient response to cancer treatment in detail, providing them with the information necessary to make a decision that might have high impact on a patient’s quality of life.