Automatic Segmentation of Large Blood Vasculature in DCE-MRI Data of Brain Tumor Using Different Clustering Algorithms

Motivation: The presence of normal large-blood-vessels(LBV) in tumor region can impact the evaluation of quantitative DCE-MRI parameters and tumor classification. Goal(s): To develop an automated framework for segmenting LBVs present within or around the tumor region using different clustering algorithms and compare their accuracy in tumor grading. Approach: LBV masks were generated using three different clustering algorithms on the DCE-MRI derived maps CBV and Slope-2. Generated tumor mask using AI tool on FLAIR images. Statistical analysis was performed. Results: Overall, k-means clustering based algorithm provided superior performance in segmentation of LBV and tumor grading in less computational time. Impact: The proposed automatic LBV segmentation algorithm can assist radiologists in objective and accurate assessment of tumor including tumor grading. This will reduce errors in tumor assessment.