Valgus and Varus Disease Detection using Image Processing

This study focuses on digitizing the measurement of intermalleolar distance (IMD), intercondylar distance (ICD) and the detection of valgus (knock knees) and varus (bowlegs) disease. IMD is the distance between two ankles while ICD is the distance between two knees. These two measurements can be used to determine if a person has a related disease or not. This paper utilizes the YOLO algorithm and Holistically-Nested edge detection to detect the subject knees, ankles and its edges. The YOLO algorithm allows the training of the system enabling it to make the detection more accurate. On the other hand, the Holistically-Nested edge detection eliminates some pre-processing steps and making it more accurate detection in natural images. The Euclidean distance or Euclidean Metrics and the use of DPI for measuring IMD and ICD have been explored in this paper. The prototype has been able to measure the IMD and ICD distances from an image and the results of it outputs the related disease. The results are verified by independent-sample t-test and it shows that the results from both methods does not have a significant difference, which means this study can produce similar results to the traditional method of measuring IMD and ICD distances.