Ultrasound image segmentation is still a challenging issue in various applications to extract the meaningful information for disease diagnosis in the athletes. Generally, ultrasound images could be suffering from some problems such as speckle, attenuation, signal dropout and shadows which make the segmentation process more complicated and inefficient. Due to these problems, traditional segmentation approaches could not be applicable. To overcome these problems, the current study proposed an automatic multilevel segmentation framework for ankle Anterior Talofibular Ligament (ATFL). This framework used the association of active contour and the particle swarm optimization method with curve evaluation and energy minimization capability to obtain the optimized segmented outcomes. It would be more efficiently detect the ATFL ligament in ultrasound images with better interpretation capability. Finally, this study presents various experimental segmented outcomes and corresponding analysis. On the basis of this analysis, the average sensitivity, specificity and accuracy of the proposed framework would be 80.73 , 96.57 and 94.12 respectively. © 2016 IEEE.