An Integrated Deep Learning Model for Pancreatic Cancer Segmentation and Classification based on CT Images
Koteswaramma Dodda, G. Muneeswari · 2025
Early and accurate diagnosis is needed for successful treatment of pancreatic cancer, as survival rates are low. Computed tomography (CT) is a commonly used diagnostic and staging tool for pancreatic cancers. The difficult task for clinicians and research is that exact segmentation and classification through medical images. On the other hand, manually segmenting volumetric CT scans is a laborious and subjective procedure. Recently the U-Net technique has demonstrated remarkable step in semantic segmentation on medical images. The crucial part is that to differentiate the tumour and non-tumour pancreatic tissues in CT images is the classification of pancreatic cancer. There have been various deep learning models put out. This study presents a unified deep learning model that can accurately partition and classify pancreatic cancer. The U-Net model was trained on annotated CT scans to precisely segment the pancreas and potential lesions, achieving accuracy 99.4%. The DenseNet-121 classifier was subsequently applied to the segmented regions to differentiate tumor and non-tumor tissue. The classification model achieved an accuracy of 99.5%. The aim of this model is to implement a dependable and precise diagnostic system that improves the performance of pancreatic cancer diagnosis.