Study on AI-assisted Multimodal Diagnosis System of Autoimmune Pancreatitis

The existing comprehensive diagnostic system for autoimmune pancreatitis (AIP) is complex, with multidimensional clinical information including morphological changes and a lack of specific biomarkers. Endoscopic ultrasound (EUS) can provide all the elements for morphological diagnosis of AIP, but the long learning curve and large observer differences make it difficult to popularize and promote. The cooperation units of the three regions in this project have found in the early stage that Klebsiella pneumoniae (KP) induced follicular helper T cells (Tfh) activation is an important mechanism of AIP, but the identification of pathogenic components of the strain and clinical validation need to be explored. We have established a national multicenter AIP queue in the early stage and extracted EUS audio-visual features to establish a scoring model, but intelligent assistance is still needed to improve efficiency. Therefore, we plan to integrate gut microbiota, Tfh activation markers, and EUS imaging features to establish an AI assisted multimodal diagnostic system for AIP. This study will collaborate across multiple centers to identify and validate the components that induce Tfh activation in KP bacterial cells, to extract EUS pancreatic ultrasound features and optimize artificial intelligence assisted diagnostic algorithms, and to establish and validate an artificial intelligence assisted multimodal diagnostic system based on clinical information, biomarkers, and EUS. The aim of this study is to provide new diagnosis and treatment evaluation methods for AIP with high accuracy, convenience, and easy promotion for clinical practice.