Analysis of metacarpophalangeal profiles by pattern recognition techniques.

RATIONALE AND OBJECTIVES The author introduces a number of techniques well known in exploratory pattern recognition that can be used for the analysis of the shape of metacarpophalangeal profiles. It is shown that the classic Q-scores must be adapted for such techniques to be applicable. METHODS A new set of scores (P-scores) describing the shape of metacarpophalangeal profiles is derived. The application of various pattern-recognition techniques that use P-scores is described, using a collection of metacarpophalangeal length measurements in patients with various pathologic conditions. RESULTS Different pattern-recognition techniques highlight different aspects of the profiles, which, when interpreted together, yield a consistent understanding of the data set and insight in individual patients' peculiarities. CONCLUSIONS The use of scale-invariant scores is imperative when the shape of profiles is to be analyzed, especially in data sets with large variations in the scale factor. Methods of pattern recognition using such scores are of potential clinical interest.