Enhanced Deep Learning Model for Diagnosis of Pancreatic Solid Lesions Through Multimodal Clinical Images

Solid lesions of the pancreas mainly include tumor and non tumor lesions. More than 90% of pancreatic tumor lesions are pancreatic cancer, which is characterized by high mortality and poor prognosis and requires surgical treatment; Non-tumor lesions of the pancreas are mainly inflammatory lesions, which usually do not require surgical treatment, but can be treated with drugs. The common ones are chronic pancreatitis and autoimmune pancreatitis, with a good prognosis. Clinically, the differential diagnosis between them is very difficult. Multi-disciplinary diagnosis and treatment of MDT makes our understanding of pancreatic diseases increasingly rich and in-depth. From disease diagnosis to preoperative evaluation and curative effect evaluation, non-invasive imaging involves almost every link under MDT mode. In view of this, improving the differential diagnosis of pancreatic solid space-occupying lesions on imaging will be more conducive to the diagnosis and treatment under MDT mode, so new technologies such as artificial intelligence should be considered. Our goal is to develop a clinically applicable artificial intelligence system, which uses multiple modes to simulate the routine clinical workflow and assist in the diagnosis of benign and malignant pancreatic solid space-occupying lesions.