Artificial intelligence for gastric cancer in endoscopy: From diagnostic reasoning to market.

Recognition of gastric conditions during endoscopy exams, including gastric cancer, usually requires specialized training and a long learning curve. Besides that, the interobserver variability is frequently high due to the different morphological characteristics of the lesions and grades of mucosal inflammation. In this sense, artificial intelligence tools based on deep learning models have been developed to support physicians to detect, classify, and predict gastric lesions more efficiently. Even though a growing number of studies exists in the literature, there are multiple challenges to bring a model to practice in this field, such as the need for more robust validation studies and regulatory hurdles. Therefore, the aim of this review is to provide a comprehensive assessment of the current use of artificial intelligence applied to endoscopic imaging to evaluate gastric precancerous and cancerous lesions and the barriers to widespread implementation of this technology in clinical routine.

ChatGPT4 Outperforms Endoscopists for Determination of Postcolonoscopy Rescreening and Surveillance Recommendations.

BACKGROUND & AIMS: Large language models including Chat Generative Pretrained Transformers version 4 (ChatGPT4) improve access to artificial intelligence, but their impact on the clinical practice of gastroenterology is undefined. This study compared the accuracy, concordance, and reliability of ChatGPT4 colonoscopy recommendations for colorectal cancer rescreening and surveillance with contemporary guidelines and real-world gastroenterology practice. METHODS: History of present illness, colonoscopy data, and pathology reports from patients undergoing procedures at 2 large academic centers were entered into ChatGPT4 and it was queried for the next recommended colonoscopy follow-up interval. Using the McNemar test and inter-rater reliability, we compared the recommendations made by ChatGPT4 with the actual surveillance interval provided in the endoscopist's procedure report (gastroenterology practice) and the appropriate US Multisociety Task Force (USMSTF) guidance. The latter was generated for each case by an expert panel using the clinical information and guideline documents as reference. RESULTS: Text input of de-identified data into ChatGPT4 from 505 consecutive patients undergoing colonoscopy between January 1 and April 30, 2023, elicited a successful follow-up recommendation in 99.2% of the queries. ChatGPT4 recommendations were in closer agreement with the USMSTF Panel (85.7%) than gastroenterology practice recommendations with the USMSTF Panel (75.4%) (P < .001). Of the 14.3% discordant recommendations between ChatGPT4 and the USMSTF Panel, recommendations were for later screening in 26 (5.1%) and for earlier screening in 44 (8.7%) cases. The inter-rater reliability was good for ChatGPT4 vs USMSTF Panel (Fleiss κ, 0.786; 95% CI, 0.734-0.838; P < .001). CONCLUSIONS: Initial real-world results suggest that ChatGPT4 can define routine colonoscopy screening intervals accurately based on verbatim input of clinical data. Large language models have potential for clinical applications, but further training is needed for broad use.

Upfront endoscopic necrosectomy or step-up endoscopic approach for infected necrotising pancreatitis (DESTIN): a single-blinded, multicentre, randomised trial.

BACKGROUND: Although the preferred management approach for patients with infected necrotising pancreatitis is endoscopic transluminal stenting followed by endoscopic necrosectomy as step-up treatment if there is no clinical improvement, the optimal timing of necrosectomy is unclear. Therefore, we aimed to compare outcomes between performing upfront necrosectomy at the index intervention versus as a step-up measure in patients with infected necrotising pancreatitis. METHODS: This single-blinded, multicentre, randomised trial (DESTIN) was done at six tertiary care hospitals (five hospitals in the USA and one hospital in India). We enrolled patients (aged ≥18 years) with confirmed or suspected infected necrotising pancreatitis with a necrosis extent of at least 33% who were amenable to endoscopic ultrasound-guided drainage. By use of computer-generated permuted block randomisation (block size four), eligible patients were randomly assigned (1:1) to receive either upfront endoscopic necrosectomy or endoscopic step-up treatment. Endoscopists were not masked to treatment allocation, but participants, research coordinators, and the statistician were. Lumen-apposing metal stents (20 mm diameter; 10 mm saddle length) were used for drainage in both groups. In the upfront group, direct necrosectomy was performed immediately after stenting in the same treatment session. In the step-up group, direct necrosectomy or additional drainage was done at a subsequent treatment session if there was no clinical improvement (resolution of any criteria of systemic inflammatory response syndrome or sepsis or one or more organ failure and at least a 25% percentage decrease in necrotic collection size) 72 h after stenting. The primary outcome was the number of reinterventions per patient to achieve treatment success from index intervention to 6 months' follow-up, which was defined as symptom relief in conjunction with disease resolution on CT. Reinterventions included any endoscopic or radiological procedures performed for necrosectomy or additional drainage after the index intervention, excluding the follow-up procedure at 4 weeks for stent removal. All endpoints and safety were analysed by intention-to-treat. This study is registered with ClinicalTrials.gov, NCT05043415 and NCT04113499, and recruitment and follow-up have been completed. FINDINGS: Between Nov 27, 2019, and Oct 26, 2022, 183 patients were assessed for eligibility and 70 patients (24 [34%] women and 46 [66%] men) were randomly assigned to receive upfront necrosectomy (n=37) or step-up treatment (n=33) and included in the intention-to-treat population. At the time of index intervention, seven (10%) of 70 patients had organ failure and 64 (91%) patients had walled-off necrosis. The median number of reinterventions was significantly lower for upfront necrosectomy (1 [IQR 0 to 1] than for the step-up approach (2 [1 to 4], difference -1 [95% CI -2 to 0]; p=0·0027). Mortality did not differ between groups (zero patients in the upfront necrosectomy group vs two [6%] in the step-up group, difference -6·1 percentage points [95% CI -16·5 to 4·5]; p=0·22), nor did overall disease-related adverse events (12 [32%] patients in the upfront necrosectomy group vs 16 [48%] patients in the step-up group, difference -16·1 percentage points [-37·4 to 7·0]; p=0·17), nor procedure-related adverse events (four [11%] patients in the upfront necrosectomy group vs eight [24%] patients in the step-up group, difference -13·4 percentage points [-30·8 to 5·0]; p=0·14). INTERPRETATION: In stabilised patients with infected necrotising pancreatitis and fully encapsulated collections, an approach incorporating upfront necrosectomy at the index intervention rather than as a step-up measure could safely reduce the number of reinterventions required to achieve treatment success. FUNDING: None.

A Preliminary Controlled Trial of Endoscopic Ultrasound-guided Fiducial Markers to Guide Pancreas Surgery.

OBJECTIVE: Endoscopic ultrasound (EUS) is routinely used for fiducial marker placement (FMP) to guide stereotactic radiation of pancreatic tumors, but EUS-FMP explicitly to guide surgery has not been studied in a prospective, controlled manner. Multipurpose EUS systems have been developed that facilitate simultaneous EUS-FMP at the time of biopsy. We aimed to evaluate the feasibility of EUS-FMP to guide pancreatic resection. METHODS: In this prospective trial, we enrolled patients with resectable pancreas masses undergoing tissue sampling and placed preloaded fiducials immediately after biopsy. Intraprocedure confirmation of carcinoma, neuroendocrine, and nonlymphomatous neoplasia by rapid on-site evaluation and lesion size <4 cm was required. The main outcomes were the feasibility and ease of preoperative placement and intraoperative detection of the markers using predefined Likert scales. RESULTS: In 20 patients, EUS-FMP was successful before planned surgery and placement was technically straightforward (Likert Scale: 9.1 ± 1.3; range: 1, most challenging to 10, most facile). Intraoperative detection was feasible and improved when compared with a pre-established comparator of 5 representing an equivalent lesion without a marker (Likert Scale: 7.8 ± 2.2; range: 1, most difficult to 10, most facile; P = 0.011). The mean tumor size on EUS was 1.7 ± 0.9 (range: 0.5 to 3.6) cm. CONCLUSION: EUS-FMP is feasible and safe for resectable pancreatic tumors before surgery and may assist in perioperative detection. Preloaded fiducials may be considered for placement at the time of initial referral for EUS-fine needle biopsy.