Automatic graph-cut based segmentation of bones from knee magnetic resonance images for osteoarthritis research

The block diagram depicts the main steps in the graph-cut based segmentation; the images shown below the diagram correspond to the main steps (original image, result of block-wise ROI discovery, GC output and the segmented image.) [Display omitted] ► In this study, a new, fully automated, content-based system is proposed for knee bone segmentation from MRI. ► The purpose of the bone segmentation is to support the discovery and characterization of imaging biomarkers for osteoarthritis. ► The segmentation algorithm includes a novel content-based, two-pass disjoint block discovery mechanism. ► This algorithm requires constructing a graph using image data followed by applying a maximum-flow algorithm. ► The results show an automatic bone detection rate of 0.99 and an average accuracy of 0.95 using 376 MR images.. In this paper, a new, fully automated, content-based system is proposed for knee bone segmentation from magnetic resonance images (MRI). The purpose of the bone segmentation is to support the discovery and characterization of imaging biomarkers for the incidence and progression of osteoarthritis, a debilitating joint disease, which affects a large portion of the aging population. The segmentation algorithm includes a novel content-based, two-pass disjoint block discovery mechanism, which is designed to support automation, segmentation initialization, and post-processing. The block discovery is achieved by classifying the image content to bone and background blocks according to their similarity to the categories in the training data collected from typical bone structures. The classified blocks are then used to design an efficient graph-cut based segmentation algorithm. This algorithm requires constructing a graph using image pixel data followed by applying a maximum-flow algorithm which generates a minimum graph-cut that corresponds to an initial image segmentation. Content-based refinements and morphological operations are then applied to obtain the final segmentation. The proposed segmentation technique does not require any user interaction and can distinguish between bone and highly similar adjacent structures, such as fat tissues with high accuracy. The performance of the proposed system is evaluated by testing it on 376 MR images from the Osteoarthritis Initiative (OAI) database. This database included a selection of single images containing the femur and tibia from 200 subjects with varying levels of osteoarthritis severity. Additionally, a full