Order of importance of anatomical risk factors in Kienböck's disease: An artificial neural network study.

Anatomical variations in bony structures around the wrist have been considered as risk factors for Kienböck's disease: ulnar variance, Nattrass index, ulnar variance/capitate height ratio and presence of a lunohamate joint. This study aimed to assess the order of importance of these variations as risk factors for Kienböck's disease. Two groups were formed: patients (n = 58) and controls (n = 235). On posteroanterior radiographs in the two groups, these risk factors were examined by four raters. After inter-rater correlation analysis, an artificial neural network was used to estimate their relative importance. All parameters showed statistically significant inter-rater correlation (p < 0.05). The artificial neural network study showed that the three most important risk factors, in descending order, were: Nattrass index, ulnar variance/capitate height ratio and negative ulnar variance. The study determined the order of importance of the anatomical risk factors for Kienböck's disease measurable on posteroanterior wrist radiographs. Although these findings seem to be useful in the diagnostic algorithm of Kienböck's disease, multivariate analysis of all measurable risk factors is still needed. The artificial neural network approach could contribute to such a comprehensive study.