Chari, Suresh T.

Background Failure of conventional imaging to detect pancreatic ductal adenocarcinoma (PDA) at its visually occult pre-diagnostic stage is a primary barrier to improving its otherwise poor rate of survival. Objective To develop and validate the Radiomics-based Early Detection MODel (REDMOD), an AI framework to identify subvisual radiomic signatures of pre-diagnostic PDA on standard-of-care CT. Designs REDMOD was trained on a multi-institutional cohort (n=969; 156 pre-diagnostic, 813 control) and tested on an independent set (n=493; 63 pre-diagnostic, 430 control), simulating a low prevalence (~1:6) early detection paradigm. The fully automated framework couples AI-driven segmentation with a heterogeneous ensemble architecture trained on a 40-feature radiomic signature derived from Synthetic Minority Over-sampling Technique (SMOTE)-balanced data. A tunable Youden Index-optimised classification threshold enables performance calibration without retraining. Validation included direct comparison with radiologists, longitudinal test–retest analysis and external specificity validation across two independent cohorts (n=539 and n=80). Results On an independent test set (n=493), REDMOD identified occult PDA (AUC 0.82; 73.0% sensitivity) at a median 475-day lead time. This represented nearly twofold higher sensitivity than radiologists (38.9%; p<0.001), which grew to nearly threefold (68.0% vs 23.0%) at >24 months lead time. REDMOD showed strong longitudinal stability (90–92% concordance) and generalisable specificity across multi-institutional (81.3%; n=539) and public (87.5%; n=80) datasets. Mechanistic analyses confirmed predictive power derived principally from multi-scale wavelet-filtered textural features (90% of selected signature), which outperformed unfiltered features (AUC 0.82 vs 0.74; p=0.007) in capturing subvisual architectural disruptions. Conclusions REDMOD is an automated, mechanistically grounded, longitudinally stable, externally validated AI that surpasses radiologists for PDA detection at its visually occult pre-diagnostic stage. These attributes position it for prospective validation in high-risk cohorts, a necessary step towards shifting the paradigm from late-stage symptomatic diagnosis to proactive pre-clinical interception.

Abstract Background: Early detection is the most potent strategy to improve survival in PDA, but its success hinges on an unresolved question: does radiologically invisible disease represent biologically early or indolent disease, or established malignancy beneath the threshold of detection? We leveraged prediagnostic CTs—incidental scans obtained 3-36 months prior to diagnosis without a focal mass—to define the relationship between imaging phenotype and tumor biology. Methods: From a multi-institutional pool of 5,046 biopsy-confirmed PDAs, we assembled a cohort of 346 prediagnostic CTs, which were independently evaluated by three radiologists using a standardized protocol. Based on majority vote, CTs were dichotomized into imaging-occult (IO) (normal pancreas) and imaging-apparent (IA) (abnormal pancreas with indirect signs such as focal atrophy, duct dilatation, etc.) groups. Tumor size and metastases (M) at diagnosis, and overall survival (OS) were compared using multivariable linear (size), logistic (M1), and Cox proportional hazards (OS) models, adjusting for lead time to diagnosis, age, sex, diabetes, CA 19-9, and surgical resection status, to isolate the independent biological relevance of the imaging phenotype. Results: IO phenotype was seen in 150 patients (mean age 66.7±11.2; 37.3% females) while 196 had IA phenotype (mean age 69.9±10.6; 49.5% females). Median lead time was 396 days (range: 90-1088) with the IA group being enriched in 3-12-month lead time (57.7% vs 33.3%; p&lt;0.001). Both groups had comparable diabetes rates (IA 33.2% vs IO 32.7%; p=1.0), median CA 19-9 (IA 193 vs IO 322 U/mL; p=0.41), and surgical resection (IA 38.3% vs IO 38.0%; p = 1.0). While unadjusted mean tumor size at diagnosis was larger in IA group (3.8 cm vs 3.2 cm), there was no difference after adjusting for clinical factors (-7.1% difference; 95% CI: -17.2, 4.2; p=0.21). Metastases were also comparable (M rate: IA 22.4% vs IO 27.3%; adjusted OR 0.97; 95% CI: 0.50, 1.88; p=0.93). Adjusted Cox models demonstrated equivalent OS (median: 18.7 vs 17.6 months; HR 1.06; p=0.66), which persisted even when stratified by surgical status (HR 1.07; 95% CI: 0.83, 1.38; p=0.61). Even in the critical 3-12 months lead time (n=163), OS remained equivalent (HR 0.89; p=0.62), and groups did not differ in either tumor size (-6.4%; 95% CI: -22.4, 12.9; p=0.49) or metastases (OR 0.48; 95% CI 0.14, 1.59; p=0.23). Conclusions: A normal-appearing pancreas on prediagnostic imaging is common yet signifies neither early nor indolent disease—it harbors PDA biologically equivalent to the imaging-apparent pancreatic phenotype. Because the disease at this stage lies beyond human perceptual capacity, AI augmentation represents the only potential path to realizing the survival benefit of early detection. Citation Format: Khurram Khaliq Bhinder, Sovanlal Mukherjee, Armin Zarrintan, Angela Ammirabile, Ahmed Khan Jadoon, Ziding Feng, Suresh T. Chari, Ajit H. Goenka. Dissociation between prediagnostic imaging phenotype and tumor biology defines the early detection challenge in pancreatic ductal adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(8_Suppl):Abstract nr LB205.

ImportanceIntraductal papillary mucinous neoplasms (IPMNs) are pancreatic cysts that can give rise to pancreatic cancer (PC). Limited population data exist on their prevalence, natural history, or risk of malignant transformation (IPMN-PC).ObjectiveTo fill knowledge gaps in epidemiology of IPMNs and associated PC risk by estimating population prevalence of IPMNs, associated PC risk, and proportion of IPMN-PC.Design, Setting, and Participants: This retrospective cohort study was conducted in Olmsted County, Minnesota. Using the Rochester Epidemiology Project (REP), patients aged 50 years and older with abdominal computed tomography (CT) scans between 2000 and 2015 were randomly selected (CT cohort). All patients from the REP with PC between 2000 and 2019 were also selected (PC cohort). Data were analyzed from November 2021 through August 2023.Main outcomes and MeasuresCIs for PC incidence estimates were calculated using exact methods with the Poisson distribution. Cox models were used to estimate age, sex, and stage–adjusted hazard ratios for time-to-event end points.ResultsThe CT cohort included 2114 patients (1140 females [53.9%]; mean [SD] age, 68.6 [12.1] years). IPMNs were identified in 231 patients (10.9%; 95% CI, 9.7%-12.3%), most of which were branch duct (210 branch-duct [90.9%], 16 main-duct [6.9%], and 5 mixed [2.2%] IPMNs). There were 5 Fukuoka high-risk (F-HR) IPMNs (2.2%), 39 worrisome (F-W) IPMNs (16.9%), and 187 negative (F-N) IPMNs (81.0%). After a median (IQR) follow-up of 12.0 (8.1-15.3) years, 4 patients developed PC (2 patients in F-HR and 2 patients in F-N groups). The PC incidence rate per 100 person years for F-HR IPMNs was 34.06 incidents (95% CI, 4.12-123.02 incidents) and not significantly different for patients with F-N IPMNs compared with patients without IPMNs (0.16 patients; 95% CI, 0.02-0.57 patients vs 0.11 patients; 95% CI, 0.06-0.17 patients; P = .62). The PC cohort included 320 patients (155 females [48.4%]; mean [SD] age, 72.0 [12.3] years), and 9.8% (95% CI, 7.0%-13.7%) had IPMN-PC. Compared with 284 patients with non-IPMN PC, 31 patients with IPMN-PC were older (mean [SD] age, 76.9 [9.2] vs 71.3 [12.5] years; P = .02) and more likely to undergo surgical resection (14 patients [45.2%] vs 60 patients [21.1%]; P = .003) and more-frequently had nonmetastatic PC at diagnosis (20 patients [64.5%] vs 130 patients [46.8%]; P = .047). Patients with IPMN-PC had better survival (adjusted hazard ratio, 0.62; 95% CI, 0.40-0.94; P = .03) than patients with non-IPMN PC.Conclusions and RelevanceIn this study, CTs identified IPMNs in approximately 10% of patients aged 50 years or older. PC risk in patients with F-N IPMNs was low and not different compared with patients without IPMNs; approximately 10% of patients with PC had IPMN-PC, and they had better survival compared with patients with non-IPMN PC.

Purpose This study aimed to compare accuracy and efficiency of a convolutional neural network (CNN)–enhanced workflow for pancreas segmentation versus radiologists in the context of interreader reliability. Methods Volumetric pancreas segmentations on a data set of 294 portal venous computed tomographies were performed by 3 radiologists (R1, R2, and R3) and by a CNN. Convolutional neural network segmentations were reviewed and, if needed, corrected (“corrected CNN [c-CNN]” segmentations) by radiologists. Ground truth was obtained from radiologists' manual segmentations using simultaneous truth and performance level estimation algorithm. Interreader reliability and model's accuracy were evaluated with Dice-Sorenson coefficient (DSC) and Jaccard coefficient (JC). Equivalence was determined using a two 1-sided test. Convolutional neural network segmentations below the 25th percentile DSC were reviewed to evaluate segmentation errors. Time for manual segmentation and c-CNN was compared. Results Pancreas volumes from 3 sets of segmentations (manual, CNN, and c-CNN) were noninferior to simultaneous truth and performance level estimation–derived volumes [76.6 cm3 (20.2 cm3), P < 0.05]. Interreader reliability was high (mean [SD] DSC between R2-R1, 0.87 [0.04]; R3-R1, 0.90 [0.05]; R2-R3, 0.87 [0.04]). Convolutional neural network segmentations were highly accurate (DSC, 0.88 [0.05]; JC, 0.79 [0.07]) and required minimal-to-no corrections (c-CNN: DSC, 0.89 [0.04]; JC, 0.81 [0.06]; equivalence, P < 0.05). Undersegmentation (n = 47 [64%]) was common in the 73 CNN segmentations below 25th percentile DSC, but there were no major errors. Total inference time (minutes) for CNN was 1.2 (0.3). Average time (minutes) taken by radiologists for c-CNN (0.6 [0.97]) was substantially lower compared with manual segmentation (3.37 [1.47]; savings of 77.9%–87% [P < 0.0001]). Conclusions Convolutional neural network–enhanced workflow provides high accuracy and efficiency for volumetric pancreas segmentation on computed tomography.

PurposeTo develop a two‐stage three‐dimensional (3D) convolutional neural networks (CNNs) for fully automated volumetric segmentation of pancreas on computed tomography (CT) and to further evaluate its performance in the context of intra‐reader and inter‐reader reliability at full dose and reduced radiation dose CTs on a public dataset.MethodsA dataset of 1994 abdomen CT scans (portal venous phase, slice thickness ≤ 3.75‐mm, multiple CT vendors) was curated by two radiologists (R1 and R2) to exclude cases with pancreatic pathology, suboptimal image quality, and image artifacts (n = 77). Remaining 1917 CTs were equally allocated between R1 and R2 for volumetric pancreas segmentation [ground truth (GT)]. This internal dataset was randomly divided into training (n = 1380), validation (n = 248), and test (n = 289) sets for the development of a two‐stage 3D CNN model based on a modified U‐net architecture for automated volumetric pancreas segmentation. Model’s performance for pancreas segmentation and the differences in model‐predicted pancreatic volumes vs GT volumes were compared on the test set. Subsequently, an external dataset from The Cancer Imaging Archive (TCIA) that had CT scans acquired at standard radiation dose and same scans reconstructed at a simulated 25% radiation dose was curated (n = 41). Volumetric pancreas segmentation was done on this TCIA dataset by R1 and R2 independently on the full dose and then at the reduced radiation dose CT images. Intra‐reader and inter‐reader reliability, model’s segmentation performance, and reliability between model‐predicted pancreatic volumes at full vs reduced dose were measured. Finally, model’s performance was tested on the benchmarking National Institute of Health (NIH)‐Pancreas CT (PCT) dataset.ResultsThree‐dimensional CNN had mean (SD) Dice similarity coefficient (DSC): 0.91 (0.03) and average Hausdorff distance of 0.15 (0.09) mm on the test set. Model’s performance was equivalent between males and females (P = 0.08) and across different CT slice thicknesses (P > 0.05) based on noninferiority statistical testing. There was no difference in model‐predicted and GT pancreatic volumes [mean predicted volume 99 cc (31cc); GT volume 101 cc (33 cc), P = 0.33]. Mean pancreatic volume difference was −2.7 cc (percent difference: −2.4% of GT volume) with excellent correlation between model‐predicted and GT volumes [concordance correlation coefficient (CCC)=0.97]. In the external TCIA dataset, the model had higher reliability than R1 and R2 on full vs reduced dose CT scans [model mean (SD) DSC: 0.96 (0.02), CCC = 0.995 vs R1 DSC: 0.83 (0.07), CCC = 0.89, and R2 DSC:0.87 (0.04), CCC = 0.97]. The DSC and volume concordance correlations for R1 vs R2 (inter‐reader reliability) were 0.85 (0.07), CCC = 0.90 at full dose and 0.83 (0.07), CCC = 0.96 at reduced dose datasets. There was good reliability between model and R1 at both full and reduced dose CT [full dose: DSC: 0.81 (0.07), CCC = 0.83 and reduced dose DSC:0.81 (0.08), CCC = 0.87]. Likewise, there was good reliability between model and R2 at both full and reduced dose CT [full dose: DSC: 0.84 (0.05), CCC = 0.89 and reduced dose DSC:0.83(0.06), CCC = 0.89]. There was no difference in model‐predicted and GT pancreatic volume in TCIA dataset (mean predicted volume 96 cc (33); GT pancreatic volume 89 cc (30), p = 0.31). Model had mean (SD) DSC: 0.89 (0.04) (minimum–maximum DSC: 0.79 −0.96) on the NIH‐PCT dataset.ConclusionA 3D CNN developed on the largest dataset of CTs is accurate for fully automated volumetric pancreas segmentation and is generalizable across a wide range of CT slice thicknesses, radiation dose, and patient gender. This 3D CNN offers a scalable tool to leverage biomarkers from pancreas morphometrics and radiomics for pancreatic diseases including for early pancreatic cancer detection.

Abstract Purpose: We have previously identified tissue methylated DNA markers (MDMs) associated with pancreatic ductal adenocarcinoma (PDAC). In this case–control study, we aimed to assess the diagnostic performance of plasma MDMs for PDAC. Experimental Design: Thirteen MDMs (GRIN2D, CD1D, ZNF781, FER1L4, RYR2, CLEC11A, AK055957, LRRC4, GH05J042948, HOXA1, PRKCB, SHISA9, and NTRK3) were identified on the basis of selection criteria applied to results of prior tissue experiments and assays were optimized in plasma. Next, 340 plasma samples (170 PDAC cases and 170 controls) were assayed using target enrichment long-probe quantitative amplified signal method. Initially, 120 advanced-stage PDAC cases and 120 healthy controls were used to train a prediction algorithm at 97.5% specificity using random forest modeling. Subsequently, the locked algorithm derived from the training set was applied to an independent blinded test set of 50 early-stage PDAC cases and 50 controls. Finally, data from all 340 patients were combined, and cross-validated. Results: The cross-validated area under the receiver operating characteristic curve (AUC) for the training set was 0.93 (0.89–0.96) for the MDM panel alone, 0.91 (95% confidence interval, 0.87–0.96) for carbohydrate antigen 19-9 (CA19-9) alone, and 0.99 (0.98–1) for the combined MDM-CA19-9 panel. In the test set of early-stage PDAC, the AUC for MDMs alone was 0.84 (0.76–0.92), CA19-9 alone was 0.87 (0.79–0.94), and combined MDM-CA19-9 panel was 0.90 (0.84–0.97) significantly better compared with either MDMs alone or CA19-9 alone (P = 0.0382 and 0.0490, respectively). At a preset specificity of 97.5%, the sensitivity for the combined panel in the test set was 80% (28%–99%) for stage I disease and 82% (68%–92%) for stage II disease. Using the combined datasets, the cross-validated AUC was 0.9 (0.86–0.94) for the MDM panel alone and 0.89 for CA19-9 alone (0.84–0.93) versus 0.97 (0.94–0.99) for the combined MDM-CA19-9 panel (P ≤ 0.0001). Overall, cross-validated sensitivity of MDM-CA19-9 panel was 92% (83%–98%), with an observed specificity of 92% at the preset specificity of 97.5%. Conclusions: Plasma MDMs in combination with CA19-9 detect PDAC with significantly higher accuracy compared with either biomarker individually.

Objective The diagnosis of autoimmune pancreatitis (AIP) is challenging. Sonographic and cross-sectional imaging findings of AIP closely mimic pancreatic ductal adenocarcinoma (PDAC) and techniques for tissue sampling of AIP are suboptimal. These limitations often result in delayed or failed diagnosis, which negatively impact patient management and outcomes. This study aimed to create an endoscopic ultrasound (EUS)-based convolutional neural network (CNN) model trained to differentiate AIP from PDAC, chronic pancreatitis (CP) and normal pancreas (NP), with sufficient performance to analyse EUS video in real time. Design A database of still image and video data obtained from EUS examinations of cases of AIP, PDAC, CP and NP was used to develop a CNN. Occlusion heatmap analysis was used to identify sonographic features the CNN valued when differentiating AIP from PDAC. Results From 583 patients (146 AIP, 292 PDAC, 72 CP and 73 NP), a total of 1 174 461 unique EUS images were extracted. For video data, the CNN processed 955 EUS frames per second and was: 99% sensitive, 98% specific for distinguishing AIP from NP; 94% sensitive, 71% specific for distinguishing AIP from CP; 90% sensitive, 93% specific for distinguishing AIP from PDAC; and 90% sensitive, 85% specific for distinguishing AIP from all studied conditions (ie, PDAC, CP and NP). Conclusion The developed EUS-CNN model accurately differentiated AIP from PDAC and benign pancreatic conditions, thereby offering the capability of earlier and more accurate diagnosis. Use of this model offers the potential for more timely and appropriate patient care and improved outcome.

Purpose of review Pancreatic cancer is the third leading cause of cancer death and with a dismal 5-year survival of 10%. Poor survival of pancreatic cancer is mostly due to its presentation and diagnosis at a late stage. The present article aims to update clinicians with recent progress in the field of early detection of pancreatic cancer. Recent findings Pancreatic cancer screening is not recommended in the general population due to its low prevalence. In this review, we discuss high-risk groups for pancreatic cancer, including inherited predisposition to pancreatic cancer, new-onset diabetes, mucinous pancreatic cyst, and chronic pancreatitis. We discuss methods of enrichment of high-risk groups with clinical models using electronic health records and biomarkers. We also discuss improvements in imaging modalities and emerging role of machine learning and artificial intelligence in the field of imaging and biomarker to aid in early identification of pancreatic cancer. Summary There are still vast challenges in the field of early detection of pancreatic cancer. We need to develop noninvasive prediagnostic validated biomarkers for longitudinal surveillance of high-risk individuals and imaging modalities that can identify pancreatic cancer early.

Abstract Background Endoscopic intervention for pancreatic fluid collections (PFCs) with disconnected pancreatic duct syndrome (DPDS) has been associated with failures and increased need for additional endoscopic and non-endoscopic interventions. The primary aim of this study was to determine the outcomes of endoscopic ultrasound (EUS)-guided transmural drainage of PFCs in patients with DPDS. Methods In patients undergoing EUS-guided drainage of PFCs from January 2013 to January 2018, demographic profiles, procedural indications and details, adverse events, outcomes, and subsequent interventions were retrospectively collected. Overall treatment success was determined by PFC resolution on follow-up imaging or stent removal without recurrence. Results EUS-guided drainage of PFCs was performed in 141 patients. DPDS was present in 57 of them (40 %) and walled-off necrosis was the most frequent type of PFC (55 %). DPDS was not associated with lower clinical success, increased number of repeat interventions, or increased time to PFC resolution. Patients with DPDS were more likely to be treated with permanent transmural plastic double-pigtail stents (odds ratio [OR] 6.4; 95 % confidence interval [CI] 2.5 – 16.5; P < 0.001). However, when stents were removed, DPDS was associated with increased PFC recurrence after stent removal (OR 8.0; 95 %CI 1.2 – 381.8; P = 0.04). Conclusions DPDS frequently occurs in patients with PFCs but does not negatively impact successful resolution. DPDS is associated with increased PFC recurrence after stent removal.