Citations

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.

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.

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.

Abstract H3K27M-mutant diffuse midline glioma (DMG) patients have no proven effective therapies beyond radiation. ONC201, a mitochondrial protease ClpP agonist, has recently demonstrated efficacy in these patients, but the mechanism behind this remains unknown. We assessed clinical outcomes, tumor sequencing, and tissue samples from patients treated in two completed multi-site clinical studies (n=71). Patients treated with ONC201 monotherapy following initial radiation but prior to recurrence (n=35) demonstrated a median overall survival of 21.7 months and a median progression-free survival of 12.2 months. Radiographic response was associated with increased expression of key tricarboxylic acid cycle-related genes in baseline tumor sequencing. Similar survival benefits were observed using in utero electroporation and orthotopic mouse models of H3K27M-DMG (median 107 versus 77 days [p=0.02] and median 97 versus 141 days [p=0.004], respectively). Integrated transcriptomic and metabolomic analysis of ONC201-treated H3K27M-DMG cells revealed downregulation of TCA cycle, glycolysis, and pyruvate metabolism genes, most notably oxoglutarate dehydrogenase (OGDH), and concomitant changes in levels of associated metabolites including ɑ-ketoglutarate (ɑ-KG). Interestingly, L-2-hydroxyglutarate (L-2HG), a known inhibitor of the Jumonji C domain family of histone lysine demethylases, was also increased. This corresponded with increases in repressive H3K27me3 in vitro and in human tumors accompanied by epigenetic downregulation of cell cycle regulation and neuro-glial differentiation genes. Genetic knockdown of OGDH recapitulated ONC201-mediated increases in H3K27me3. In contrast, genetic knockdown of ClpP or lactate dehydrogenase A or overexpression of L-2-hydroxyglutarate dehydrogenase (L-2HGDH), but not D-2HGDH, abrogated ONC201-mediated increases in H3K27me3, demonstrating that ONC201 induces the production of L-2HG in H3K27M-DMG cells leading to increased H3K27me3. Overall, our data demonstrates the efficacy of ONC201 in H3K27M-mutant DMG and supports ONC201 as the first monotherapy to improve outcomes in patients with H3K27M-mutant DMG for whom few therapeutic options currently exist. Mechanistically, ONC201 disrupts integrated metabolic and epigenetic pathways and reverses pathognomonic H3K27me3 reduction.

Abstract Patients with H3K27M-mutant diffuse midline glioma (DMG) have no proven effective therapies. ONC201 has recently demonstrated efficacy in these patients, but the mechanism behind this finding remains unknown. We assessed clinical outcomes, tumor sequencing, and tissue/cerebrospinal fluid (CSF) correlate samples from patients treated in two completed multisite clinical studies. Patients treated with ONC201 following initial radiation but prior to recurrence demonstrated a median overall survival of 21.7 months, whereas those treated after recurrence had a median overall survival of 9.3 months. Radiographic response was associated with increased expression of key tricarboxylic acid cycle–related genes in baseline tumor sequencing. ONC201 treatment increased 2-hydroxyglutarate levels in cultured H3K27M-DMG cells and patient CSF samples. This corresponded with increases in repressive H3K27me3 in vitro and in human tumors accompanied by epigenetic downregulation of cell cycle regulation and neuroglial differentiation genes. Overall, ONC201 demonstrates efficacy in H3K27M-DMG by disrupting integrated metabolic and epigenetic pathways and reversing pathognomonic H3K27me3 reduction. Significance: The clinical, radiographic, and molecular analyses included in this study demonstrate the efficacy of ONC201 in H3K27M-mutant DMG and support ONC201 as the first monotherapy to improve outcomes in H3K27M-mutant DMG beyond radiation. Mechanistically, ONC201 disrupts integrated metabolic and epigenetic pathways and reverses pathognomonic H3K27me3 reduction. This article is featured in Selected Articles from This Issue, p. 2293

Background Many hormones display distinct circadian rhythms, driven by central regulators, hormonal bioavailability, and half-life. A set of 11-oxygenated C19 steroids (11-oxyandrogens) and pregnenolone sulfate (PregS) are elevated in congenital adrenal hyperplasia and other disorders, but their circadian patterns have not been characterized. Participants and methods Peripheral blood was collected every 2 h over 24 h from healthy volunteer men (10 young, 18–30 years, and 10 older, 60–80 years). We used mass spectrometry to quantify 15 steroids, including androstenedione (A4), testosterone (T), 11β-hydroxy- and 11-ketotestosterone (11OHT, 11KT),11β-hydroxy- and 11-ketoandrostenedione (11OHA4, 11KA4), and 4 ∆5-steroid sulfates. Diurnal models including mesor (rhythm adjusted median), peak, and nadir concentrations, acrophase, and amplitude were computed. Results 11OHA4 followed a rhythm similar to cortisol: acrophase 8:00 h, nadir 21:00 h and were similar in young and old men. 11KT had similar diurnal patterns, but the peak was lower in older than in young men, as was the case for A4. All four steroid sulfates were higher in young vs older men. PregS and 17-hydroxypregnenolone sulfate (17OHPregS) showed sustained elevations between 8:00 and 18:00 h, and nadirs around midnight, while DHEAS and AdiolS displayed minimal diurnal variations. All 4 11-oxyandrogens correlated tightly with cortisol (r from 0.54 for 11OHT to 0.81 for 11OHA4, P < 0.0001 for all), but very weakly with T, supporting their adrenal origin and ACTH governance. Conclusions 11-Oxyandrogens, PregS, and 17OHPregS display distinct circadian and age variations, which should be accounted for when used as clinical biomarkers.

Introduction . Primary cardiac sarcoma (PCS) has a poor prognosis compared to other sarcomas due to late presentation, challenging resection, incidence of metastases, and limited efficacy of systemic therapies. Methods . A medical record search engine was queried to identify patients diagnosed with PCS from 1992 to 2017 at the University of Michigan. Results . Thirty-nine patients with PCS had a median age of 41 years (range 2–77). Common histologies were angiosarcoma (AS, 14), high-grade undifferentiated pleomorphic sarcoma (UPS, 10), and leiomyosarcoma (LMS, 5). Sites of origin were left atrium (18), right atrium (16), and pericardium (5). AS was the most common right-sided tumor; UPS was more common on the left. Eighteen patients presented with metastases involving lung (10), bone (7), liver (5), and brain (4). Twenty-five patients underwent resection, achieving 3 R 0 resections. Patients received a median of 2 (1–6) systemic therapies. Median overall survival (OS) was 12.1 months (range 0–79). Median OS was 14.0 months and 8.2 months in patients who did or did not undergo resection, respectively ( p = 0.018 ). Brain metastases occurred in 12 (31%) patients, 9 (75%) of whom had left heart tumors, at a median of 8.5 months (range 0–75) from diagnosis. Median OS was 5.6 months (range 0–30) after the diagnosis of brain metastases. Conclusions . PCS portends a poor prognosis, because of difficulty in obtaining complete resection of sarcoma, advanced stage at diagnosis, and high risk of brain metastases. Providers should be aware of the increased risk of brain metastases and consider brain imaging at diagnosis and follow-up.