Skip to main content

MEDICAL IMAGE ANALYSIS FOR CANCER TREATMENT MONITORING AND TUMOR ATLAS FORMATION

Final Report Summary - MICAT (MEDICAL IMAGE ANALYSIS FOR CANCER TREATMENT MONITORING AND TUMOR ATLAS FORMATION)

Cancer disease is one of the leading causes of death in humans for developed nations such as European Union. During monitoring of cancer patients undergoing therapy, quantitative analysis of the change in tumor is a substantial task for an accurate assessment of the disease. Within the current clinical environment, physicians usually have to go through many manual outlining, visual matching and comparison over the patient scans, which decrease efficiency of the clinical workflow. Computational tools specifically designed for cancer treatment monitoring have the potential to enhance this process significantly. MICAT project's objectives were to develop new computational tools for image segmentation, alignment, and measurement of tumor response from baseline and follow-up Magnetic Resonance Imaging (MRI) scans of patients with brain tumors who are undergoing radio-surgery. In addition, MICAT project led to re-integration of the PI, Dr. Gozde Unal, to European Research Area, and facilitated establishment of her independent research group VpaMed at the host institution, Sabanci University, Istanbul, Turkey. By building a strong capacity on medical image computing, she trained young scientists, including several PhD students, as well as setting up an inter-disciplinary team with strong clinical collaborations. The project team included radio-oncology and radiology experts from two hospitals: Anadolu Medical Center, and Hacettepe University Hospital.
A new segmentation tool for extracting surfaces of 3D tumor volumes from brain MRI was developed based on the cellular-automata framework, starting with a single line on a 2D MR plane, which followed the radiological practice of tumor maximum diameter evaluation. The algorithm delineates the gross tumor volume and necrotic regions of the brain tumors on contrast enhanced T1-weighted MRI. It was also extended to multimodal tumor segmentation in order to delineate the edema regions and non-enhancing content of the tumor tissue over T1, T2, and FLAIR image volumes. The algorithm was among top 3 of the state-of-the-art algorithms in the Live Challenge on Multimodal BRain Tumor Segmentation (BRaTS) at MICCAI 2012.
New algorithms for alignment of brain tumor volumes at baseline and follow-up MRI were developed. First, a rigid registration method based on automatically extracted anatomical landmarks was designed. Next, a deformable registration was utilized to map the follow-up volume onto the coordinate space of the baseline volume. The latter facilitated exploration of new local tumor response criteria. In addition to computation of global measures of tumor response that are based on tumor volume by World Health Organization (WHO), and tumor maximum in plane diameter by Response Evaluation Criteria in Solid Tumors (RECIST), we proposed measures, based on continuum mechanics, of Lagrange strain tensor invariants computed from deformation field between the baseline and follow-up tumor volumes. In a preliminary study, these local measures correlated well with the clinical findings of the disease progression, stability or regression. The constructed graphical user interface with the tools from the MICAT project, has the potential to be further developed into a sophisticated software platform for increased efficiency in quantitative and qualitative analysis of brain tumors in both stages of radio-surgery planning and tumor response follow-up.
Website: http://vpa.sabanciuniv.edu/phpBB2/vpa_views.php?s=4&serial=39
final1-micat.png