Multimodal Imaging Diagnosis and Decision Aid System for Hepatic Echinococcosis Based on Image Omics and Vision Macromodel

• Automatic recognition/Efficient diagnosis of hepatic Echinococcosis: Development of a deep learn-based AI diagnostic tool aimed at improving the differentiation of hepatic echinococcosis from other liver diseases such as liver cysts, liver abscesses, and other hepatic cystic space occupying lesions. The tool will utilize generative adversarial networks and polarized self-attention algorithms to effectively identify and classify hepatic echinococcosis to make up for the uneven medical resources and shortage of professional physicians in the western region
• Differential diagnosis of specific types of hepatic echinococcosis: To explore the use of multimodal imaging combined with deep learning methods to distinguish CL type, CE1 type and hepatic cyst of hepatic echinococcosis. This research will apply DINOv2 medical image segmentation algorithm and deep learning technology to accurately identify cases with relatively unevenly distributed and complex data.
• Prediction of postoperative recurrence of hepatic echinococcosis: Transfer learning and Deep-SVDD algorithm are used to predict the risk of postoperative recurrence of hepatic echinococcosis, providing an effective solution for unbalanced sample size data sets. In addition, by integrating the data on the medical big data platform, an AI-based postoperative recurrence prediction model was established

Inclusion criteria

• Patients with complete original images in CT, ultrasound, and MRI dcim formats
• Patients with liver hydatid confirmed by pathology after operation
• Patients with complete clinical data preservation

Exclusion criteria

• Patients with poor quality imaging data
• Patients with incomplete clinical data
• CE4 and CE5 liver hydatid patients diagnosed by imaging alone without surgical treatment