An automated content-based segmentation framework: Application to MR images of knee for osteoarthritis research

To effectively diagnose and monitor the treatment of diseases such as osteoarthritis, the segmentation, processing and analysis of mass volumes of medical images is gaining high importance. In this paper, a new fully automated content-based segmentation framework is proposed. The framework is designed to be compatible with a wide variety of segmentation techniques. To this end, a novel content-based two-pass block discovery mechanism is proposed to provide full automation for image segmentation. The proposed framework uses both training and local image data and disjoint block-wise image scanning to achieve ROI and background block discovery. The detected object and background blocks are then used to initialize and support the segmentation process. The effectiveness of the proposed framework is demonstrated by performing automatic segmentation of the femur and tibia bones in knee osteoarthritis MR images with 96% accuracy. Experimental results are provided which show the effectiveness of the proposed framework.