Key points:
  1. Accurate preoperative prediction of the CRS type is necessary for predicting postoperative outcomes and administering personalized treatment
  2. We combined semantic segmentation with classification network for sinus region segmentation and differentiating between ECRS and NECRS on preoperative CT
  3. The segmentation model could accurately segment the sinus region in CT images
  4. The classification model could effectively distinguish between patients with ECRS and NECRS based on preoperative sinus CT images
  5. Labeling each patient to build a dataset for classification may be more reliable than labeling each medical image.