Computer assisted bone age assessment

Abstract

Bone age assessment based on the radiological examination of the left hand and wrist is a procedure frequently performed to evaluate the growth of pediatric patients. In this thesis we examined studies conducted on the skeletal age analysis and we developed a computerized system for automatic bone age analysis. Firstly we examined two different clinical methods; The Greulich and Pyle method and the Tanner and Whitehouse (TW2) method. We chose the TW2 method to develop a computerized bone age system because it is more suitable for the computer analysis than the Greulich and Pyle method. We developed the web-based TW2 skeletal age calculation software. We then introduced steps that are necessary to transform this manual clinical method into a fully automated system. We explained preprocessing methods that are needed to make a radiograph fit for analysis and steps to find regions of interest and assign them TW2 stages. Finally we developed a computer program that assigns TW2 stages to a certain region of interest (the middle phalanx of the third finger) in the left hand. In this method we trained all TW2 stage groups for middle phalanx of the third finger and produced mean images for each TW2 stage. These mean images are used to determine the TW2 stage of queried image. We use the correlation between a mean image and a target image as an indicator of which stage should be assigned to the target.Our results show that 70% of all test images of the middle phalanx of the third finger have been assigned the correct TW2 stage by our method. We believe that the same method could be used to assign TW2 stages to other regions of interest in the hand.