Optimizing Wrist Splint Fitting Parameters Through Artificial Intelligence Analysis

Introduction The current method for determining the appropriate wrist splint size in the clinical setting relies on measuring wrist circumference, but this approach often fails to ensure optimal fit. This study evaluates additional hand features using 3-dimensional (3D) scanned data and Artificial Intelligence (AI) to improve the fit of pre-fabricated wrist splints. We hypothesize that wrist and forearm widths can provide a more accurate fitting than wrist circumference alone. Materials and methods We recruited 54 healthy volunteers to be scanned. Each volunteer was fitted with a standard wrist brace (Short Arm Brace, Ossur, Iceland), and 3D data from their hands were collected using an infrared-based 3D scanner (Einscan Pro, Shining3D, China). The 3D scanned data were then analyzed to identify and measure 14 distinct hand features. To explore the relationship between these hand features and the optimal splint size, we generated a categorical correlation map. This map identified hand features that were most strongly correlated with splint size categories (small, medium, large). Subsequently, we developed a classification algorithm to predict the appropriate splint size based on the correlated hand features. We utilized three different machine learning models for this purpose: Extreme Gradient Boosting (XGB) Classifier, RandomForestClassifier, and Support Vector Classifier (SVC). Each of these classifiers was trained and evaluated to determine their accuracy and effectiveness in predicting the correct splint size. Results Wrist width showed the highest classification accuracy (91%) for both the XGB Classifier and RandomForestClassifier. The measurements including hand wrist width, mid-forearm width, and hand crease line width also performed well with the XGB Classifier, achieving an accuracy of 90%. The SVC showed consistent performance across various feature sets, with the highest accuracy of 81% for the measurements. Overall, these findings suggest that wrist width is the most predictive feature for splint size classification, with additional features providing minimal enhancement. Conclusions Artificial intelligence, combined with 3D scanning, can accurately predict wrist splint size from a single image acquisition, enabling contactless, personalized fitting. This approach can improve patient outcomes by enhancing the fit of prefabricated splints.