Skeletal bone age assessment from epiphysis/metaphysis of phalanges using Hausdorff distance

Abstract

The work explains an efficient method for skeletal bone age assessment (BAA) from the epiphysis/metaphysis of phalanges using Hausdorff distance as the measure for classification. The method was based on the Tanner and Whitehouse (TW2) method of BAA. The system guaranteed accurate BAA for the age range 0-10 years for both male and female. The input image was initially preprocessed and 8 EMROI joints were identified and cropped, which were segmented using difference of the Gaussian (DoG) method. 14 features were extracted from each EMROI and stored as feature vector. In addition, three distance parameters were extracted to compute the closeness of the epiphysis and metaphysis of the fifth finger middle phalanx as a measure of fusion. Finally, classification of the image into one of the ten bone age classes was done by measuring the Hausdorff distance between the trained feature vector and the feature set of the test image. The assigned age class (Class A – J), was then mapped onto the final bone age. The performance of the system was evaluated with the help of diagnosis done by two skilled radiologists. The system was very robust, achieving an accuracy rate of 98%, specificity rate 99%, precision rate 88%, and recall rate 90% when partitioning the dataset into 120 train images and 100 test images. When partitioned into 160 train images and 60 test images, the system scored 99% for accuracy, 99% for specificity, 95% for precision and 94% for recall. For the partition with 180 train images and 40 test images, the proposed method showed 100% results in all the four performance metrics. Key words: Skeletal bone age assessment (BAA), TW2, radiograph, EMROI, epiphysis, metaphysis, Hausdorff distance, classification.