Automatic bone age measurement using computerized image analysis.

In 1992 we described a computer-assisted method for assigning Tanner-Whitehouse RUS skeletal maturity scores to hand-wrist radiographs. An operator positions each epiphysis in turn beneath a video camera, views the image on the computer screen and corrects the position of the radiograph by matching to templates of the TW stages displayed on the screen. The process is then automatic; the computer, not the operator, rates the bone. The image is digitized and then represented by a large number of mathematical coefficients. These coefficients are then compared to those generated by each stage of the TW standards, and the closest match is sought. Since the comparison is quantitative the system produces continuous stage scores instead of the old discrete ones such as B, C, D, etc. Thus in longitudinal data a much smoother progression of skeletal maturity scores with age is achieved. The reliability of the computer-assisted skeletal age score (CASAS) is considerably greater than that of the usual manual method. Differences between duplicate readings of a bone by a single observer average about 0.25 stage, and reach 1.0 stage or more only in about 3% of instances, compared with 15-20% characteristic of manual ratings.