Anterior Cruciate Ligament Segmentation: Using Morphological Operations with Active Contour

Among the ligaments responsible in maintaining the structural integrity of knee joint, anterior cruciate ligament (ACL) injury is most commonly diagnosed. Recent advancement in clinical imaging technology has led to wide employment of magnetic resonance imaging (MRI) in such injury assessment. However, the visual assessment conducted with these images often requires the boundaries of selected structures to be manually traced using computer software. Such interpretation is often time consuming and subjective as it is based on the radiologist's opinion and past experiences. In this study, a semi-automatic ACL segmentation program that utilized both morphological operations and active contour is proposed. It takes advantage of the ACL's unique shape and orientation within MR images to carry out the segmentation. Among 111 PD-weighted images segmented, the proposed program was capable of achieving an overall sensitivity, specificity and Dice coefficient of 43.3 % ±± 14.0 %, 99.4 % ±± 0.3 %, and 0.381 ±± 0.091 respectively. Although these values indicated low performance produced by the proposed program, the results from this study did prove its feasibility in providing an objective and reproducible ACL segmentation. Thus, with necessary improvements implemented, this program can be deployed clinically to facilitate ACL injury diagnosis.