Banerjee, Imon

Background Pericarditis may be associated with significant morbidity and mortality and can be a clinically challenging entity to diagnose. There is a clinical need for automated artificial intelligence (AI) algorithms to act as a screening aid to assist with the early and accurate detection of pericarditis. Methods We developed p-Cal, a browser-based pericarditis risk calculator that fuses the patient clinical data and 12-lead ECG image to estimate risk for pericarditis. The ECG encoder was designed based on a pretrained MedCLIP Vision Transformer and random forest classifier was used for the clinical tabular data. These two modalities were fused at the decision-level using a meta classifier. Model reasoning for both modalities was assessed using advanced explainable AI techniques—attention maps from the transformer model were used to interpret ECG images, while SHapley Additive exPlanations (SHAP) force plots were applied to analyse the tabular data. Results A total of 6508 patients (mean age 54.8±16.1 years, 50.9% female) were used for training and internal validation of the AI model, of whom 902 patients had a confirmed diagnosis of pericarditis (13.8%) and the remainder were normal controls. On 1302 hold-out test patients, the Fusion Model demonstrated the best discriminatory performance, achieving an area under the curve (AUC) of 0.89, which surpassed the ECG (AUC = 0.85) and the clinical data only models (AUC = 0.74) performance. On external validation in the Medical Information Mart for Intensive Care (MIMIC) Database, the fusion model achieved an AUC of 0.81. Conclusion A late fusion AI model integrating tabular clinical electronic medical record data and ECG signal has robust predictive capability for pericarditis as a screening tool.

Background Immune checkpoint inhibitors (ICI) improve cancer outcomes but are associated with increased atherosclerotic cardiovascular risk. Opportunistic coronary artery calcium (CAC) assessment from routine non-gated computed tomography (CT) using artificial intelligence (AI) may enable rapid cardiovascular risk stratification in this population. We evaluated the association between AI-derived CAC (AI-CAC) scores and 1-year myocardial infarction (MI) and all-cause mortality in patients receiving ICI therapy. Methods We conducted a retrospective multicenter study across three Mayo Clinic sites including patients who underwent chest CT within 3 months prior to ICI initiation between 2011 and 2022. Patients with prior coronary revascularization were excluded. AI-CAC scores derived from non-gated CT scans were categorized as 0, 1–99, 100–399, and ≥400. Cox regression analyses assessed associations with 1-year MI and all-cause mortality. Results Among 1,873 patients (mean age 66 years; 43% women), lung cancer was most common (50.3%), and 91% had at least one cardiovascular risk factor. Higher CAC burden was associated with increased 1-year MI risk (overall p=0.0032). AI-CAC was not associated with all-cause mortality (p=0.79), likely reflecting high cancer-related mortality (50.7% at 1 year). Only 15% of patients were receiving statins despite one-third having CAC >100. Conclusions Opportunistic AI-CAC assessment from routine CT scans provides modest prognostic value for MI risk among patients receiving ICI therapy and may identify opportunities to optimize preventive cardiovascular care. Elevated CAC should prompt risk factor modification rather than influence ICI treatment decisions. Prospective studies are warranted to evaluate clinical implementation and outcomes.

BACKGROUND: Accurate assessment of left ventricular outflow tract (LVOT) gradients is critical for hypertrophic cardiomyopathy management, yet Doppler-based measurements are technically demanding and require expertise. The objective of this work was to develop a multi-view deep learning model capable of classifying LVOT obstruction (>20 mm Hg) using routine 2-dimensional echocardiographic windows without reliance on Doppler imaging. METHODS: We trained and externally validated a cross-attention-based video-to-video fusion framework that integrated EchoPrime-derived video representations from 3 standard transthoracic echocardiographic views to classify LVOT gradients. RESULTS: Training was performed on a derivation cohort (N=1833) from a tertiary care system in the United States, with model performance evaluated on an internally held-out test set (N=275) and a Korean external validation cohort (N=46). Single-view baselines showed limited discrimination (external area under the receiver operating curves, 0.47–0.70). Conversely, the domain-specific foundational model (EchoPrime) achieved superior single-view performance (area under the receiver operating curves, 0.75–0.80 internal; 0.79–0.83 external), highlighting the importance of echo-specific pretraining and temporal modeling. The proposed multi-view fusion further enhanced predictive performance, with the late fusion model reaching an area under the receiver operating curve of 0.84 on the external cohort with significant population-shift. CONCLUSIONS: These results suggest LVOT physiology is encoded in routine 2-dimensional imaging and can be leveraged for clinically relevant gradient classification without Doppler input. The proposed artificial intelligence-guided strategy demonstrates substantial cost savings compared with the screen-all approach. By integrating complementary spatial-temporal information across multiple views, our approach generalizes robustly across populations and may enable real-time decision support, extend LVOT assessment to portable or resource-limited settings, and complement Doppler-based evaluation for longitudinal hypertrophic cardiomyopathy management.

PURPOSE To evaluate open-source large language models (LLMs) for extracting cancer-specific phenotypic data, benchmark their performance against GPT4 models, and assess the impact of fine-tuning with training data sizes. METHODS Open-source LLMs (Mistral, LLaMa, MAMBA, BioMistral) were evaluated in zero-/one-shot and fine-tuned setups against GPT4-turbo/GPT4o to extract the cancer presence, progression, response, and metastatic sites from radiology impressions of patients with solid tumors treated at Dana-Farber Cancer Institute. Performance metrics (accuracy, precision, recall, F1-score) were computed. McNemar's odds ratio (OR), measuring which model is more likely to be correct when they disagree, was computed with 95% CI. Statistical significance was assessed using the alpha of .000139. RESULTS This study included 2,623 patients (25,273 radiology impressions). In zero-/one-shot settings, GPT4-turbo/GPT4o outperformed open-source LLMs. However, fine-tuned open-source LLMs achieved higher F1-scores than GPT4 models. Compared with the best-performing GPT4 model, fine-tuned Mistral0.2-7.3B (OR, 0.27 [95% CI, 0.20 to 0.36]; P < .00001), Mistral0.3-7.3B (OR, 0.26 [95% CI, 0.19 to 0.36]; P < .00001), LLaMa2-6.7B (OR, 0.30 [95% CI, 0.22 to 0.40]; P < .00001), LLaMa3.1-8B (OR, 0.37 [95% CI, 0.28 to 0.48]; P < .00001), and MAMBA-2.8B (OR, 0.32 [95% CI, 0.24 to 0.42]; P < .00001) showed significantly better performance in ascertaining disease progression. Performance was consistently better for inferring overall response, any evidence of cancer, and sites of metastases, with no significant differences among fine-tuned open-source LLMs. Fine-tuning gains plateaued at 25% of training data (5,718 impressions) and remained comparable at 5% (1,144 impressions). CONCLUSION Open-source LLMs, when fine-tuned using labeled data, can effectively automate the ascertainment of key radiophenotypic variables using only the impression section of radiology reports, without the full report text. Their consistent performance in small training sets suggests that these models may provide a scalable approach for phenotypic characterization of patients with cancer in real-world clinical settings.

e16017 Background: Pancreatic ductal adenocarcinoma (PDAC) is the most common form of pancreatic cancer and remains one of the deadliest malignancies worldwide, largely due to late-stage diagnosis. Early-stage PDAC is frequently asymptomatic, limiting opportunities for timely detection and intervention. We aimed to develop prognostic models using routinely acquired pre-diagnostic clinical data to identify individuals at high risk for PDAC prior to clinical manifestation. Methods: We conducted a retrospective study of patients treated at Mayo Clinic sites in Rochester, MN; Phoenix, AZ; and Jacksonville, FL between January 2011 and February 2023. The cohort included 1,151 patients with PDAC and 3,745 noncancer controls. CT scans were standardized to axial abdominal acquisitions with slice thickness between 0.8–3.0 mm. Radiology reports were de-identified, normalized, and parsed into structured sections (Indications, Findings, Impression), with an additional pancreas-specific section aggregating pancreas-related sentences. Demographic and clinical variables, e.g., age, sex, race, etc., were encoded using a templated sentence representation. All textual inputs were embedded using a Sentence-BERT model. A multimodal deep learning survival model was developed to integrate CT imaging, radiology report embeddings, and demographic features. The model was trained using a negative log-likelihood loss to estimate patient-specific PDAC risk scores. Model performance was assessed using the concordance index (C-index). Internal validation included patients from MN and AZ sites, while external validation used the FL cohort. Results: We evaluated unimodal (imaging, text, and demographics) and multimodal inputs. Fusion of imaging, radiology report, and demographic features achieved the highest performance, with internal and external validation C-indices of 0.885 (95% CI, 0.856–0.908) and 0.751 (95% CI, 0.719–0.783), respectively. Subgroup analysis by tumor location demonstrated differential performance. For pancreatic head tumors, internal and external C-indices were 0.916 (95% CI, 0.841–0.960) and 0.778 (95% CI, 0.730–0.823). For pancreatic body tumors, C-indices were 0.954 (95% CI, 0.897–0.997) internally and 0.929 (95% CI, 0.897–0.953) externally. For pancreatic tail tumors, internal and external C-indices were 0.903 (95% CI, 0.830–0.974) and 0.789 (95% CI, 0.689–0.874). Conclusions: A multimodal deep survival model leveraging pre-diagnostic CT imaging, radiology report embeddings, and demographic data can effectively identify individuals at high risk for PDAC prior to clinical diagnosis. Differences in performance were observed across pancreatic head, body, and tail tumors. This approach highlights the potential of opportunistic cancer risk stratification from routine imaging to enable earlier detection and targeted intervention strategies.

Abstract Background Accurate assessment of left ventricular outflow tract (LVOT) gradients is critical for hypertrophic cardiomyopathy (HCM) management, yet Doppler-based measurements are technically demanding and require expertise. Objective To develop a multi-view deep learning model capable of classifying LVOT obstruction (> 20mmHg) using routine 2D echocardiographic windows without reliance on Doppler imaging. Methods We trained and externally validated a cross-attention-based video-to-video fusion framework that integrated EchoPrime-derived video representations from three standard transthoracic echocardiographic views to classify LVOT gradients. Results Training was performed on a derivation cohort (N = 1833) from a tertiary care system in the United States, with model performance evaluated on an internal held-out test set (N = 275) and a Korean external validation cohort (N = 46). Single-view baselines showed limited discrimination (external AUROCs 0.47–0.70). Conversely, domain-specific foundational model (EchoPrime) achieved superior single-view performance (AUROCs 0.75–0.80 internal; 0.79–0.83 external), highlighting the importance of echo-specific pretraining and temporal modeling. The proposed multi-view fusion further enhanced predictive performance, with the late fusion model reaching an AUROC of 0.84 on the external cohort with significant population-shift. Conclusions These results suggest LVOT physiology is encoded in routine 2D imaging and can be leveraged for clinically relevant gradient classification without Doppler input- proposed AI-guided strategy demonstrates substantial cost savings compared with the screen-all approach. By integrating complementary spatial-temporal information across multiple views, our approach generalizes robustly across populations and may enable real-time decision support, extend LVOT assessment to portable or resource-limited settings, and complement Doppler-based evaluation for longitudinal HCM management.

Abstract Artificial intelligence (AI)-scores estimated from digital mammograms predict future breast cancer (BC) risk. MIRAI is a deep learning BC risk model that provides a continuous 5-year risk of overall BC from four screening full field digital mammograms. Common germline genetic variation in the form of a polygenic risk score (BC-PRS) is associated with increased BC risk and may improve MIRAI’s 5-year risk prediction. Our goal was to develop an updated MIRAI 5-year risk model that incorporates the BC-PRS (MIRAI+PRS) and to evaluate its discriminatory accuracy and calibration for both overall and invasive BC compared to MIRAI alone. We developed the MIRAI+PRS model by multiplying each woman’s 5-year MIRAI risk estimate by their relative risk based on their BC-PRS relative to the population mean. We evaluated the models within the Mayo Clinic Biobank mammography cohort, comprised of 12,307 women without a prior history of BC; 176 invasive and 250 overall BC were diagnosed within 5 years. MIRAI was estimated on screening mammograms closest to enrollment but at least 6 months prior to BC. Discriminatory accuracy, assessed by C-statistic, was high and similar for MIRAI+PRS vs. MIRAI models, for overall BC and invasive BC (Table). Calibration assessed by observed to expected (O/E) ratios was also similar for MIRAI+PRS compared to MIRAI predictions for BC outcomes (Table), although there was improvement in decile-specific O/E ratios across the lowest risk deciles (&lt;1.67%) for MIRAI+PRS. Calibration for invasive cancer was poor for MIRAI with or without PRS. In summary, the MIRAI+PRS risk model did not result in significant difference of discriminatory accuracy or overall calibration compared to the MIRAI model, but there was evidence for improved calibration for women with 5-year risk below 1.67%. For invasive BC, the model had poor calibration regardless of whether PRS was included, underscoring the importance of training AI models for BC outcomes that are associated with a clinical intervention. Citation Format: Christopher G. Scott, Peter Kraft, Imon Banerjee, Ramon Correa Medero, Aaron D. Norman, Fergus J. Couch, Karla Kerlikowske, Stacey J. Winham, Celine M. Vachon. Development and evaluation of a MIRAI 5-year risk model with a breast cancer polygenic risk score [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2779.

Abstract Background Artificial intelligence (AI)–based, mammography breast cancer risk prediction models show improved discriminatory accuracy relative to clinical risk models. However, data on their calibration are limited. This study compared model performance of 3 clinical breast cancer risk models: Gail, Tyrer–Cuzick v8, and Breast Cancer Surveillance Consortium v3 with the MIRAI AI-risk model. Methods Digital mammograms were ascertained from a screening mammography cohort of 12 308 women within the Mayo Clinic Biobank with 250 incident breast cancers (176 invasive) within 5 years. We predicted 5-year breast cancer risk, estimated discriminatory accuracy (C-index), and calibration (observed-to-expected ratio) of both overall and invasive breast cancer and compared estimates using bootstrapping approaches. Results MIRAI demonstrated similar or improved discriminatory accuracy of overall breast cancer (C-index = 0.71, 95% confidence interval [CI] = 0.68 to 0.74) and invasive breast cancer (C-index = 0.71, 95% CI = 0.67 to 0.75) compared with clinical models (overall breast cancer: C-index = 0.59-0.68; invasive breast cancer: C-index = 0.60-0.68). MIRAI’s calibration for risk of overall breast cancer (observed-to-expected ratio = 0.96, 95% CI = 0.85 to 1.08) was improved compared with Gail (observed-to-expected ratio = 1.22, 95% CI = 1.07 to 1.38) and Breast Cancer Surveillance Consortium (observed-to-expected ratio = 1.38, 95% CI = 1.22 to 1.56) but similar to TC with volumetric percent density and polygenic risk score (observed-to-expected ratio = 0.99, 95% CI = 0.87 to 1.13). However, for low-risk women (approximately 50%), MIRAI overestimated risk of overall breast cancer. MIRAI also overestimated risk of invasive breast cancer across the risk spectrum (observed-to-expected ratio = 0.68, 95% CI = 0.58 to 0.78), whereas clinical models had good calibration (observed-to-expected ratio = 0.86–0.99). Conclusion MIRAI demonstrated stronger discriminatory accuracy than clinical models for 5-year overall and invasive breast cancer risk prediction but overestimated risk for both breast cancer endpoints. AI-based risk models should consider discriminatory accuracy and calibration for invasive cancer before implementation.