A Screening Method for Cervical Myelopathy Using Machine Learning to Analyze a Drawing Behavior

Abstract

Abstract Early detection of cervical myelopathy (CM) is important for a favorable outcome, as its prognosis is poor if left untreated. We developed a screening method for CM using machine learning to analyze a drawing behavior. A total of 38 patients with CM and 66 healthy volunteers were enrolled. Using a stylus pen, they traced three different shapes displayed on a tablet device. During the exercise, writing behaviors, such as the coordinates, velocity, and pressure of the stylus tip, along with drawing time were recorded. From these data, features related to the drawing pressure and time of each shape and combination of shapes were used as training data for the support vector machine, a machine learning algorithm. To evaluate the accuracy, a receiver operating characteristic curve was generated, and the area under the curve (AUC) was calculated. Models with triangular wave forms tended to be most accurate, and the best triangular wave model identified patients with and without CM with 76% sensitivity and 76% specificity, yielding an AUC of 0.80. Our model was able to classify CM with high accuracy and could be applied to the development of disease screening systems useful outside the hospital setting.