Rectal Cancer Tumor Image Segmentation Based on CNN Image Feature Recognition

Aiming at the problems of complex gray scale distribution, different local contrast, artifacts, and noise in CT images, this study proposes a new tumor segmentation method for rectal cancer, which employs a CNN model with a specific sequential input strategy to preprocess the images, and by comparing it with a random sequential input strategy, we demonstrate that the proposed method achieves a correctness rate of 70% in complex tumor slices. In addition to having a direct impact on the field of medical imaging, this research provides valuable insights and lessons learned for the wide application of deep learning in medical image processing.