Implementation and statistical evaluation of computer assisted TW2 method for bone age assessment

Abstract

The most commonly used method for bone age assessment is based on a single x-ray of the hand and wrist. The bones in the x-ray are compared to the bones of a standard atlas, usually "Greulich and Pyle (G&P)". A more complex method also based on hand x-rays is the "Tanner-Whitehouse (TW2)" method, which relies on the systematic evaluation of the maturity of all the bones in the hand and wrist. In this study, first we implemented the computer assisted TW2 method, then we compared this method with reference to widely used method of G&P using the criteria of accuracy and speed, and lastly we studied how learning and practice affects speed of bone age assessment. We used 50 "bone age" radiographs of the left hand and wrist performed in a large hospital. data were analyzed using the "method comparison" statistical technique. 20% of the radiographs were then re-analyzed to assess intra-obsever variation. The 95% confidence interval for the difference between the two methods was -1.84 to 1.32 years. Intra-observer variation was greater for the G&P method than for the TW2 method (95% confidence limits, -0.77 to 0.97 vs -0.45 to 0.37). The spedd of computer based TW2 was close to G&p (1.7 min vs 0.7 min) and increased wiyh practice. Since both methods take reasonable amount of time, computerized TW2 method should be preferred for higher performance in bone age assessment.|Keywords: Bone age; Greulich and Pyle; Computerized Tanner and Whitehouse.