Wallace, Michael B.

Abstract: Background Artificial intelligence (AI)-assisted colonoscopy reduces costs in colorectal cancer screening, but its incremental value relative to non-AI resect-and-discard (RD) strategies for diminutive polyps (≤ 5 mm) remain unclear. We aimed to compare the clinical, economic, and environmental outcomes of AI-assisted and non-AI RD strategies with standard of care (SoC). Methods In our Markov model simulating 6 million individuals aged ≥45 years undergoing screening colonoscopy, 4 strategies were evaluated: SoC, non-AI RD, CADe-assisted RD (CADe-RD), and CADe+CADx-assisted RD (CADe+CADx-RD). The primary outcome was total quality-adjusted life years (QALYs), with secondary outcomes including cost and carbon dioxide (CO₂) emissions. Results Non-AI RD achieved the lowest total cost ($36.6B) and carbon emissions (0.90 million kg CO₂), while yielding the highest QALYs (51.37 million), although differences in QALYs across strategies were small (<0.01%). CADe-RD and CADe+CADx-RD remained cost-saving relative to SoC ($36.9B and $37.4B vs $37.8B, respectively) but were associated with higher costs than non-AI RD. CADe modestly increased adenoma detection without improving QALYs, while the addition of CADx increased costs and emissions and resulted in slightly lower QALYs. Sensitivity analysis showed that CADe+CADx-RD matched or exceeded CADe-RD QALYs only within a limited region of high diagnostic performance. At currently reported CADx performance levels, CADe+CADx-RD yielded lower QALYs and higher costs than CADe-RD. CADe+CADx-RD achieved non-inferiority in QALYs relative to CADe-RD only within a limited region of high diagnostic performance. Conclusions In our non-deterministic model, non-AI RD provided the optimal balance of clinical benefit, cost savings, and lowest environmental impact.

Artificial intelligence (AI) is transforming gastrointestinal (GI) endoscopy into a more standardized, data-driven, and workflow-integrated field. Advances in computer-assisted detection (CADe), diagnosis (CADx), quality assessment (CAQ), natural language processing (NLP), and multimodal deep learning have expanded AI applications across colonoscopy, upper endoscopy, endoscopic ultrasound (EUS), ERCP, cholangioscopy, and capsule endoscopy. These systems have demonstrated improvements in lesion detection, procedural quality assessment, workflow efficiency, and diagnostic support. However, current evidence remains largely focused on surrogate outcomes rather than patient-centered clinical benefits, while challenges related to generalizability, explainability, regulatory oversight, automation bias, and workflow integration continue to limit widespread adoption. Future progress will depend on prospective real-world validation, diverse datasets, explainable AI frameworks, and careful integration of human–AI interaction into clinical practice. Overall, AI is evolving from a supportive adjunct into an increasingly integrated component of gastrointestinal endoscopy with the potential to improve procedural quality, diagnostic consistency, and clinical efficiency.

Abstract Despite growing interest in AI for pancreatic cyst management, no prior study has systematically investigated how lesion characteristics influence model behavior or provided a comprehensive explainability analysis in IPMN malignancy risk stratification. We present the first multi-center study integrating explainable AI with uncertainty quantification to evaluate how cyst type, size, and location affect deep learning performance in IPMN malignancy prediction. Our retrospective study analyzed 170 IPMNs from seven centers using a radiomics-deep learning fusion model. Cases were stratified by dysplasia grade, IPMN type, size, and location; with model interpretability assessed using SHAP, LIME, Grad-CAM, and uncertainty quantification. Overall accuracy was 67.1%, with model uncertainty lower for correct versus incorrect predictions (0.72 vs 0.78, p < 0.001). On subgroup analysis: Low-Risk lesion accuracy exceeded High-Risk (81.3% vs. 42.9%, p < 0.001), BD accuracy was inversely correlated with cyst size while MD remained stable, and body and whole-pancreas cysts stratified more accurately than tail cysts (77.8%, 76.6% vs. 47.8%). AI models face significant, previously uncharacterized performance drops when evaluating complex high-risk IPMNs. Our study shows that integrating uncertainty quantification can successfully flag unreliable predictions, enabling a selective-prediction framework where clinicians can confidently rely on AI for low-risk, low-uncertainty lesions while deferring high-uncertainty cases to expert review.

INTRODUCTION: Colorectal cancer remains a leading cause of cancer associated death in the United States and colonoscopy the primary screening strategy for prevention. Rates of adenomatous and serrated neoplasia detection are inversely associated with postcolonoscopy colorectal cancer. This crucial quality metric depends on accurate ascertainment of colorectal neoplasia findings from both endoscopy and histopathology records. We aimed to assess the feasibility of a random forest machine learning model to rapidly and accurately categorize colorectal neoplasia from electronic health record data. METHODS: A retrospective cohort study compared neoplasia detection rates among individuals undergoing colonoscopy at a large academic institution to develop a rule-based algorithm to categorize colorectal neoplasia from endoscopy reports and pathology systematized nomenclature of medicine – clinical terms. This cohort provided a large training set to develop a natural language processing system using a random forest approach to automatically classify unstructured pathology findings into adenoma, serrated, or advanced neoplasms. This system was manually validated through an independent holdout set. RESULTS: The training set comprised 35,953 unstructured pathology reports with matched systematized nomenclature of medicine – clinical terms from 95,188 unstructured colonoscopy reports. The final model was assessed on an independent holdout set of 337 manually annotated procedures obtaining an area under the receiver operating characteristic curve of 0.997 (confidence interval [CI] 0.994–1), 0.99 (CI 0.98–1), and 0.99 (CI 0.98–0.99) for prediction of adenoma, serrated, and advanced lesions, respectively. DISCUSSION: The random forest-based hybrid natural language processing system for classification of colonoscopy results was both accurate and explainable. NLP combined with effective machine learning algorithms can provide a scalable strategy for colonoscopy quality monitoring.

Abstract Pancreatic cancer, projected to become the second-deadliest malignancy in Western societies by 2030, requires urgent innovations in early detection and risk stratification. Intraductal papillary mucinous neoplasms (IPMNs) represent critical precursor lesions, but current clinical guidelines demonstrate suboptimal accuracy in malignancy prediction, leading to either unnecessary surgeries or missed opportunities for early intervention. In this paper, we present Cyst-X, an artificial intelligence (AI) framework that accurately predicts IPMN malignant transformation using multicenter magnetic resonance imaging (MRI) data. Unlike most previous approaches that rely on computed tomography (CT), our method capitalizes on the superior soft tissue contrast of MRI, allowing for more precise identification of subtle imaging biomarkers. We developed and validated deep learning models on 723 T1-weighted and 738 T2-weighted MRI scans from 764 patients among seven international institutions, demonstrating significantly superior performance (AUC=0.82) compared to current clinical (Kyoto) guidelines (AUC=0.75) and expert radiologists. AI-based imaging features correlate strongly with clinically recognized malignancy markers, providing potential biologically relevant insights. This approach holds promise to significantly refine clinical decision-making, reduce unnecessary surgeries, and better identify high-risk IPMN patients for timely intervention. Our approach integrates a novel pancreas segmentation network with robust classification models that identify subtle imaging biomarkers associated with malignancy risk. Importantly, we demonstrate that these models retain high performance in a privacy-preserving federated learning setting, where institutions collaboratively train AI models without exchanging patient data to address key regulatory and ethical barriers. We publicly release the Cyst-X dataset--the first large-scale, multi-center pancreatic cyst MRI collection--to accelerate research in this field. This study addresses a critical clinical need while establishing technical foundations for privacy-preserving AI in radiology that could transform pancreatic cancer management through earlier intervention and reduced unnecessary procedures. The dataset can be accessed at https://osf.io/74vfs/, and the source code for our deep learning segmentation and classification models is available at https://github.com/NUBagciLab/Cyst-X.

This review explores how multimodal foundation models (MFMs) are transforming biliary tract cancer (BTC) research. BTCs are aggressive malignancies with poor prognosis, presenting unique challenges due to difficult diagnostic methods, molecular complexity, and rarity. Importantly, intrahepatic cholangiocarcinoma (iCCA), perihilar cholangiocarcinoma (pCCA), and distal bile duct cholangiocarcinoma (dCCA) represent fundamentally distinct clinical entities, with iCCA presenting as mass-forming lesions amenable to biopsy and targeted therapies, while pCCA manifests as infiltrative bile duct lesions with challenging diagnosis and primarily palliative management approaches. MFMs offer potential to advance research by integrating radiological images, histopathology, multi-omics profiles, and clinical data into unified computational frameworks, with applications tailored to these distinct BTC subtypes. Key applications include enhanced biomarker discovery that identifies previously unrecognizable cross-modal patterns, potential for improving currently limited diagnostic accuracy—though validation in BTC-specific cohorts remains essential—accelerated drug repurposing, and advanced patient stratification for personalized treatment. Despite promising results, challenges such as data scarcity, high computational demands, and clinical workflow integration remain to be addressed. Future research should focus on standardized data protocols, architectural innovations, and prospective validation studies. The integration of artificial intelligence (AI)-based methodologies offers new solutions for these historically challenging malignancies. However, current evidence for BTC-specific applications remains largely theoretical, with most studies limited to proof-of-concept designs or related cancer types. Comprehensive clinical validation studies and prospective trials demonstrating patient benefit are essential prerequisites for clinical implementation. The timeline for evidence-based clinical adoption likely extends 7–10 years, contingent on successful completion of validation studies addressing current evidence gaps.

Abstract Endoscopic mucosal resection (EMR) techniques for large (≥20 mm) nonpedunculated colorectal polyps (LNPCPs) have expanded with the introduction of ablation and cold EMR. This study assessed adverse events (AEs) for the newer EMR techniques, including cold EMR, compared with hot EMR. We conducted a secondary analysis of four prospective multicenter studies of consecutive patients with LNPCPs undergoing EMR from 2019 to 2024. The primary outcome was serious AEs (SAEs) with cold and hot EMR. Secondary outcomes included SAEs in the hot EMR subgroups (no ablation [hEMR], margin ablation [hEMR-m], margin and base ablation [hEMR-mb]). 1762 patients (mean age 65.8; 1890 LNPCPs) were included: 522 cold and 1368 hot EMRs (368 hEMR, 770 hEMR-m, 230 hEMR-mb). SAEs were higher with hot EMR (4.7%, 95%CI 3.6%–5.9%) vs. cold EMR (1.9%, 95%CI 0.9%–3.5%), also for the subgroups of hEMR (6.0%, 95%CI 3.8%–8.9%), hEMR-m (3.9%, 95%CI 2.6%–5.5%), and hEMR-mb (5.2%, 95%CI 2.7%–8.9%). Serious postendoscopic bleeding (PEB) was numerically higher with hot EMR (2.3%, 95%CI 1.6%–3.3%) vs. cold EMR (1.3%, 95%CI 0.5%–2.7%), also for the subgroups of hEMR (3.0%, 95%CI 1.5%–5.3%), hEMR-m (1.9%, 95%CI 1.1%–3.2%), and hEMR-mb (2.6%, 95%CI 1.0%–5.6%). Perforation, intraprocedural and post-procedural, was numerically higher with hot EMR (1.2%, 95%CI 0.7%–2.0%) vs. cold EMR (0.2%, 95%CI 0.0%–1.1%). hEMR-m and hEMR-mb with clipping had lower rates of serious and overall PEB than no clipping. Cold EMR demonstrated lower rates of SAEs, serious PEB, and perforation compared with hot EMR. Perforation and mortality occurred almost exclusively after hot EMR. Hot EMR with margin +/− base ablation did not increase SAEs compared with hot EMR without ablation.

ABSTRACTBackgroundPancreatic cysts are increasingly discovered on imaging studies performed for unrelated conditions. Currently, surveillance of these lesions poses a substantial burden on patients, and health care recourses. We hypothesized that individuals with small and stable cysts have a diminutive risk of progressing to high‐grade dysplasia (HGD) or pancreatic cancer (PC) that is similar to that in the general population.MethodsThis nested PACYFIC‐study is a collaboration among 44 centers in Europe and Northern‐America, and investigates the risk of HGD and PC for different cyst sizes and growth rates in participants without baseline worrisome features (WF) or high‐risk stigmata (HRS).ResultsOf the 2369 PACYFIC participants, 975 met the inclusion criteria, with a mean age of 67 years (SD 13) and 65% being female. Of these, 438 individuals (45%) had a baseline small cyst size (< 15 mm), and 885 (91%) individuals had a slow growth rate (< 2.5 mm/year). During a median follow‐up of 45 months (IQR 27), 20 individuals (2.1%) developed HGD, or PC. Individuals with small cysts had a 1.5‐fold lower risk of developing WF or HRS (hazard ratio [HR] 0.7 [0.5–1.0], p = 0.03) than those with larger cysts but a similar risk of developing HGD or PC (p > 0.05). Slow growth was protective against the development of WF or HRS (HR 0.4 [0.2–0.6], p < 0.001) and HGD or PC (HR 0.04 [95% CI 0.02–0.12], p < 0.001). Individuals with small, stable sized cysts without baseline WF or HRS did not have a higher risk of HGD or PC than the general population (standardized incidence ratio [SIR] 1.13 [95% CI 0.01–6.30]).ConclusionCyst size < 15 mm and growth rate < 2.5 mm/year appear to be “reassuring” features associated with a negligible risk of developing WF or HRS and HGD or PC. For cysts with these characteristics—and without baseline WF or HRS—less intensive surveillance (than currently recommended) or even cessation may be appropriate.

Abstract Distinguishing high-risk intraductal papillary mucinous neoplasms (IPMNs), pancreatic cysts requiring surgery, from low-risk lesions remains a clinical challenge, often resulting in unnecessary procedures due to limited specificity of current methods. While radiomics and deep learning (DL) have been explored for pancreatic cancer, cyst-level malignancy risk stratification of IPMNs remains untapped. We conducted a multi-institutional study (seven centers, 359 T2W MRI images) to assess the feasibility of AI for predicting IPMN dysplasia grade using cyst-level image features. We developed and compared 2D and 3D radiomics-only, deep learning (DL)-only, and radiomics-DL fusion models, using expert radiologist scoring as a baseline reference. Model performance was evaluated using held-out test data. The radiomics-DL fusion model showed the highest discriminatory ability on the test set (AUC 0.692), outperforming the radiomics-only model (AUC 0.665). Expert accuracy varied widely (37.4%-66.7%). The fusion model integrating deep learning and radiomics features from routine T2W MRI (AUC: 0.692) demonstrates potential for objective, cyst-level risk stratification of IPMNs in a multi-center cohort, outperforming both radiomics-only models and expert radiologists. While performance requires improvement for standalone clinical use, this approach offers a scalable, non-invasive method to potentially improve diagnostic accuracy and reduce unnecessary surgical interventions.

Objectives: New methods are needed to detect pancreatic ductal adenocarcinoma (PDAC) earlier to improve outcomes. We previously reported that a panel of protein N-glycosylation traits (NGTs) discriminated PDAC from healthy controls with an area under the curve (AUC) of 0.81–0.88. However, it remained unclear whether this panel accurately differentiates PDAC from other benign pancreatic disorders. Our study aims to evaluate the performance of the NGT panel in combination with CA19-9 in a diverse cohort, including PDAC cases, healthy controls, and controls with benign pancreatic disorders. Methods: Protein N-glycosylation profiles were determined in plasma samples using an in-house developed mass spectrometry assay. CA19-9 levels were measured using a routine immunoassay test. Results of total plasma NGTs and CA19-9 were evaluated separately as well as in combination. Logistic regression was performed to calculate odds ratios (ORs), AUC, sensitivity, and specificity to determine the performance of NGTs and CA19-9 in distinguishing PDAC from controls. Results: In total, 221 individuals were included: 45 (20.4%) with PDAC, and 176 (79.6%) controls (53 healthy and 123 with benign pancreatic disease). The AUC for differentiating PDAC from the total control cohort based on the combination of the NGT panel and CA19-9 was 0.94 (95% CI, 0.90–0.97), with a sensitivity of 0.89 (95% CI, 0.78–0.98) and specificity of 0.86 (95% CI, 0.81–0.91). Comparison of PDAC cases with healthy controls only resulted in an AUC of 0.96 (95% CI, 0.93–0.99), with a sensitivity of 0.84 (95% CI, 0.73–0.93) and specificity of 0.98 (95% CI, 0.94–1.00). Conclusions: Both plasma NGTs and CA19-9 distinguish PDAC from a diverse control cohort. The accuracy further improves when these readouts are combined, showing promise for future early detection methods.