Ray casting segmentation

Customer: N/A

Summary: In this study, we developed a segmentation algorithm that uses ray casting and boundary point detection. The resolution of the object boundary is determined by the number of rays cast, which is a variable that depends on the shape of the target region. For example, we found that using 8 rays is sufficient to segment the left ventricle with an average Dice similarity coefficient of 85%. However, the data collected from the rays resulted in a high degree of class imbalance (up to 90%). To address this problem, we applied ensemble-based classifiers such as AdaBoost.M2, RUSBoost, UnderBagging, SMOTEBagging, and SMOTEBoost for boundary detection. The proposed algorithm was tested on both a cardiac MRI dataset from the University of York and a brain tumor dataset from Southern Medical University. The highest Dice similarity coefficients for heart and brain tumor segmentation were 86.5 ± 6.9% and 89.5 ± 6.7%, respectively. The segmentation time for a heart image was 4.1 ± 2.3 ms and 20.2 ± 23.6 ms for 8 and 64 rays, respectively. Brain tumor segmentation took 5.1 ± 1.1 ms and 16.0 ± 3.0 ms for 8 and 64 rays, respectively. Overall, the proposed algorithm is fast and highly accurate for convex and closed objects and has the potential to be scaled to include different boundary detection techniques working in parallel.

Collaborators: Alejandro Frangi (University of Leeds, Leeds, United Kingdom), Olga Gerget (Tomsk Polytechnic University, Tomsk, Russia)

Project type: Research

Media: Conference paper