Research on the Classification Method of Urolithiasis Composition Based on Machine Learning

This paper aims to investigate how to accurately classify urinary calculi components by CT images. The image segmentation is completed by grayscale adjustment of the two-dimensional image and region growth of the three-dimensional image. Grayscale and energy features of the region of interest are extracted in three dimensions. In this paper, the three different models trained by the EasyEnsemble ensemble algorithm are connected in series to achieve layer-by-layer classification of unbalanced samples. The experimental results show that this method can effectively analyze the components of simple calcium oxalate monohydrate, anhydrous uric acid, carbonated apatite and dihydrate calcium oxalate in urinary tract stones, and the AUCs reach AUC=0.93, AUC=1.00, AUC= 0.86, AUC=0.76, respectively, indicating that this method can better perform multi-classification of unbalanced samples.