Background There are several types of pancreatic mass, so it is important to distinguish between them before treatment. Artificial intelligence (AI) is a mathematical technique that automates learning and recognition of data patterns. This study aimed to investigate the efficacy of our AI model using endoscopic ultrasonography (EUS) images of multiple types of pancreatic mass (pancreatic ductal adenocarcinoma [PDAC], pancreatic adenosquamous carcinoma [PASC], acinar cell carcinoma [ACC], metastatic pancreatic tumor [MPT], neuroendocrine carcinoma [NEC], neuroendocrine tumor [NET], solid pseudopapillary neoplasm [SPN], chronic pancreatitis, and autoimmune pancreatitis [AIP]). Methods Patients who underwent EUS were included in this retrospective study. The included patients were divided into training, validation, and test cohorts. Using these cohorts, an AI model that can distinguish pancreatic carcinomas from noncarcinomatous pancreatic lesions was developed using a deep-learning architecture and the diagnostic performance of the AI model was evaluated. Results 22 000 images were generated from 933 patients. The area under the curve, sensitivity, specificity, and accuracy (95 %CI) of the AI model for the diagnosis of pancreatic carcinomas in the test cohort were 0.90 (0.84–0.97), 0.94 (0.88–0.98), 0.82 (0.68–0.92), and 0.91 (0.85–0.95), respectively. The per-category sensitivities (95 %CI) of each disease were PDAC 0.96 (0.90–0.99), PASC 1.00 (0.05–1.00), ACC 1.00 (0.22–1.00), MPT 0.33 (0.01–0.91), NEC 1.00 (0.22–1.00), NET 0.93 (0.66–1.00), SPN 1.00 (0.22–1.00), chronic pancreatitis 0.78 (0.52–0.94), and AIP 0.73 (0.39–0.94). Conclusions Our developed AI model can distinguish pancreatic carcinomas from noncarcinomatous pancreatic lesions, but external validation is needed.

Endoscopic ultrasonography (EUS) is an essential diagnostic tool for various types of pancreatic diseases such as pancreatic tumors and chronic pancreatitis; however, EUS imaging has low specificity for the diagnosis of pancreatic diseases. Artificial intelligence (AI) is a mathematical prediction technique that automates learning and recognizes patterns in data. This review describes the details and principles of AI and deep learning algorithms. The term AI does not have any definite definition; almost all AI systems fall under narrow AI, which can handle single or limited tasks. Deep learning is based on neural networks, which is a machine learning technique that is widely used in the medical field. Deep learning involves three phases: data collection and annotation, building the deep learning architecture, and training and ability validation. For medical image diagnosis, image classification, object detection, and semantic segmentation are performed. In EUS, AI is used for detecting anatomical features, differential pancreatic tumors, and cysts. For this, conventional machine learning architectures are used, and deep learning architecture has been used in only two reports. Although the diagnostic abilities in these reports were about 85–95%, these were exploratory research and very few reports have included substantial evidence. AI is increasingly being used for medical image diagnosis due to its high performance and will soon become an essential technique for medical diagnosis.

Background/Aims: Endoscopic ultrasound-guided tissue acquisition (EUS-TA) is essential for the diagnosis of pancreatic cancer. The feasibility of comprehensive genomic profiling (CGP) using samples obtained by EUS-TA has been under recent discussion. This study aimed to evaluate the utility of EUS-TA for CGP in a clinical setting.Methods: CGP was attempted in 178 samples obtained from 151 consecutive patients with pancreatic cancer at the Aichi Cancer Center between October 2019 and September 2021. We evaluated the adequacy of the samples for CGP and determined the factors associated with the adequacy of the samples obtained by EUS-TA retrospectively.Results: The overall adequacy for CGP was 65.2% (116/178), which was significantly different among the four sampling methods (EUS-TA vs. surgical specimen vs. percutaneous biopsy vs. duodenal biopsy, 56.0% [61/109] vs. 80.4% [41/51] vs. 76.5% [13/17] vs. 100.0% [1/1], respectively; p=0.022). In a univariate analysis, needle gauge/type was associated with adequacy (22 G fine-needle aspiration vs. 22 G fine-needle biopsy [FNB] vs. 19 G-FNB, 33.3% (5/15) vs. 53.5% (23/43) vs. 72.5% (29/40); p=0.022). The sample adequacy of 19 G-FNB for CGP was 72.5% (29/40), and there was no significant difference between 19 G-FNB and surgical specimens (p=0.375).Conclusions: To obtain adequate samples for CGP with EUS-TA, 19 G-FNB was shown to be the best in clinical practice. However, 19 G-FNB was not still sufficient, so further efforts are required to improve adequacy for CGP.