Recent advances in PET, PET-CT, and MRI-PET image segmentation techniques

1280 Learning Objectives 1. To present current PET image segmentation techniques and recent advances. 2. To assess the accuracy and usability of current PET image segmentation methods. 3. To present PET-CT and MRI-PET image co-segmentation techniques. 1. Introduction 2. Background on PET imaging 3. Challenges of PET image segmentation a. Noise and Resolution related issues b. Discontinuous boundaries c. Large variability of pathologies 4. Evaluation of Segmentation Methods a. Surrogate truth creation b. Phantom based studies c. Commonly used quantitative measures i. Region based: Dice similarity coefficient (DSC), TPVF, and FPVF ii. Boundary based: Haussdorf distance (HD) 5. Semi-automated Segmentation Methods a. Region growing based b. Intensity based (gradient) c. Random Walk d. Graph Cut e. Fuzzy connectivity 6. Fully Automated Segmentation Methods a. Machine learning classification/clustering methods i. k-nearest neighbor, support vector machine, artificial neural network, and spectral clustering ii. Optimal thresholding value selection b. Hybrid Co-Segmentation using fused PET-CT and MRI-PET images i. Graph cut, Fuzzy connectivity, and Random walk based methods Summary i. Reviewed the state-of-the-art PET image segmentation and recent technical advances ii. Gave a general background of the challenges of PET image segmentation as well as the evaluation metrics used commonly in the literature iii. Discussed PET segmentation based on different anatomical regions iv. Recent advances in multi-modal image co-segmentation algorithm applied to PET-CT and MRI-PET hybrid images Research Support This research is supported by the Center for Infectious Disease Imaging (CIDI), the Intramural Program of the National Institutes of Allergy and Infectious Diseases (NIAID), and the National Institutes of Bio-imaging and Bioengineering (NIBIB) at the National Institutes of Health (NIH).