Kelly Brennan

All Publications

• Reducing diverse sources of noise in ventricular electrical signals using variational autoencoders EXPERT SYSTEMS WITH APPLICATIONS Ruiperez-Campillo, S., Ryser, A., Sutter, T. M., Deb, B., Feng, R., Ganesan, P., Brennan, K. A., Rogers, A. J., Kolk, M. Z. H., Tjong, F. V. Y., Narayan, S. M., Vogt, J. E. 2026; 300

  View details for DOI 10.1016/j.eswa.2025.130185

  View details for Web of Science ID 001621410100010

• The impact of preoperative antiobesity medications on weight loss in adolescents undergoing metabolic and bariatric surgery - a COSMIC study. Surgery for obesity and related diseases : official journal of the American Society for Bariatric Surgery Chinn, J. O., Shacker, M., Brennan, K. A., Kochis, M., Stetson, A., Bizimana, C., Rodrigues de Oliveira Filho, J., Hornick, M. A., Pratt, J. S., Abu El Haija, M., Griggs, C. 2025

  Abstract

  While new medications are transforming the management of obesity, their association with outcomes in adolescents undergoing metabolic and bariatric surgery (MBS) is not clear.The objective was to determine how preoperative prescription of antiobesity medications (AOMs) is associated with postoperative weight loss after MBS.The study was conducted using data from 3 academic children's hospitals, spanning the period from March 2013 to September 2024.This is a retrospective review in which demographics, obesity-related diseases, preoperative and postoperative weight and body mass index (BMI) were compared between patients who were treated preoperatively with topiramate or glucagon-like peptide-1 receptor agonists (GLP-1RAs) and those who were not. Statistical analyses included Wilcoxon rank-sum, Pearson's χ2, and Fisher's exact tests, plus 1:1 propensity score matching and multivariable linear regression sensitivity models adjusting for time-to-surgery.Of 324 patients, 22 were treated with topiramate and 30 with a GLP-1RA. Rates of obesity-related diseases were similar. Patients on GLP-1RA lost weight from first consultation to surgery (-2% BMI), while those on no medication gained (+1% BMI) and those on topiramate remained stable (0%, P = .023). There was no difference in weight/BMI at the time of surgery; however, patients pretreated with medications lost less weight than those not taking medications at 6 months (no medications: -20% BMI reduction; GLP-1RA: -18%; topiramate: -17%, P = .017) and 12 months (no medications -23% BMI reduction, GLP-1RA -15%, topiramate -17%, P = .015). From initial consultation to 12 months after surgery, the differences in weight loss between groups were not significant (P = .072).Preoperative exposure to topiramate or GLP-1RA was associated with less postoperative weight loss, despite similar starting weights/BMIs. Total weight loss from consultation through 12 months did not differ significantly between groups. These findings raise important questions regarding the use and timing of obesity management medications in relation to surgery for adolescents.

  View details for DOI 10.1016/j.soard.2025.10.014

  View details for PubMedID 41353013

• Deep Learning-Based Continuous QT Monitoring Identifies High-Risk Prolongation Events After Class III Antiarrhythmic Initiation Rogers, A., Ansari, R., Bandyopadhyay, S., Trivedi, R., Brennan, K., Ganesan, P., Perino, A., Ashley, E., Wang, P., Perez, M., Ouyang, D., Narayan, S. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4370942

  View details for Web of Science ID 001613807100048

• Development of Personalized Myocardial Surface Mesh Models with LGE Scar Integration: a Pipeline for Machine Learning and Digital Twins Liu, X., Qayyum, A., Ganesan, P., Bandyopadhyay, S., Somani, S., Brennan, K., Wang, P., Niederer, S., Narayan, S., Rogers, A. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4367687

  View details for Web of Science ID 001613944300010

• Abstract 4367773: Predicting Peak Heart Rate from Resting 12-Lead ECGs in Patients Undergoing Stress Testing using Deep Learning Liu, X., Bandyopadhyay, S., Ganesan, P., Somani, S., Brennan, K., Karius, A., Baykaner, T., Perino, A., Wang, P., Ashley, E., Perez, M., Narayan, S., Rogers, A. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4367773

  View details for Web of Science ID 001613944300042

• AI-based prediction of mortality in patients with ventricular tachycardia Bandyopadhyay, S., Sadri, S., Brennan, K., Ganesan, P., Clopton, P., Ruiperez-Campillo, S., Peralta, E., Sillett, C., Rogers, A., Narayan, S. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4370585

  View details for Web of Science ID 001613876200024

• Identifying optimum ECG features to predict sudden cardiac arrest at varying time points before the event Bandyopadhyay, S., Ganesan, P., Brennan, K., Ruiperez-Campillo, S., Ansari, R., Clopton, P., Perino, A., Wang, P., Ashley, E., Perez, M., Narayan, S., Rogers, A. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4369502

  View details for Web of Science ID 001613825200023

• Novel Foundation Models for Detecting and Generating Text Reports of Atrial Fibrillation from 12-lead ECGs in a Large Registry Ganesan, P., Peralta, E., Ruiperez-Campillo, S., Bandyopadhyay, S., Rogers, A., Chang, H., Brennan, K., Sillett, C., Clopton, P., Perino, A., Niederer, S., Narayan, S. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4370248

  View details for Web of Science ID 001613838800021

• Automated End-to-End Framework for Extracting Raw ECG Waveforms and ST Segment Values from ECG Reports and Predicting ST Elevation by Machine Learning Ganesan, P., Liu, X., Bandyopadhyay, S., Ansari, R., Somani, S., Brennan, K., Karius, A., Baykaner, T., Perino, A., Wang, P., Ashley, E., Perez, M., Narayan, S., Rogers, A. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4371591

  View details for Web of Science ID 001613816800012

• Non-Contact Magnetocardiography Localizes Atrial Foci as Accurately as High-Resolution Contact ECG Brennan, K., Bandyopadhyay, S., Ganesan, P., Ansari, R., Somani, S., Liu, X., Baykaner, T., Perino, A., Wang, P., Narayan, S., Rogers, A. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4358583

  View details for Web of Science ID 001613911000027

• Longitudinal Evaluation of Anti-Arrhythmic Drug Use to Predict Hospitalization or Death in Patients with Ventricular Tachycardia Sadri, S., Brennan, K., Bandyopadhyay, S., Ganesan, P., Desai, Y., Peralta, E., Feng, R., Sillett, C., Ruiperez-Campillo, S., Wang, P., Clopton, P., Rogers, A., Narayan, S. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4370694

  View details for Web of Science ID 001613779500004

• Large Language Models Detect Ventricular Tachycardia Recurrence in Clinical Notes and Enable Prediction of Clinical Outcomes at Scale Sadri, S., Brennan, K., Bandyopadhyay, S., Desai, Y., Ganesan, P., Peralta, E., Feng, R., Sillett, C., Ruiperez-Campillo, S., Wang, P., Clopton, P., Rogers, A., Narayan, S. LIPPINCOTT WILLIAMS & WILKINS. 2025

  View details for DOI 10.1161/circ.152.suppl_3.4366827

  View details for Web of Science ID 001613801100038

• Timing of Antiobesity Medications and Adolescent Metabolic and Bariatric Surgery. JAMA surgery Chinn, J. O., Shacker, M., Brennan, K. A., Esquivel, M. M., Pratt, J. S. 2025

  View details for DOI 10.1001/jamasurg.2025.4430

  View details for PubMedID 41123888

  View details for PubMedCentralID PMC12547670

• Does the MBSAQIP Bariatric Surgical Risk/Benefit Calculator Accurately Predict Weight Loss in Adolescents? Obesity surgery Kochis, M. A., Chinn, J. O., Nzenwa, I. C., Brennan, K. A., Pratt, J. S., Griggs, C. L. 2025

  Abstract

  BACKGROUND: The Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) online calculator incorporates individual patient data to predict weight loss up to 1year after MBS, but it was derived from an adult database and has not been validated in younger cohorts. This study evaluates the accuracy of this calculator for adolescent MBS patients and explores patient factors which may be associated with prediction inaccuracy.METHODS: We include patients age≤21 who underwent laparoscopic sleeve gastrectomy at two major academic institutions from 2013 to 2023. Data were stratified between patients age<18 and 18-21. The calculator's predictions were compared to actual weight loss values at 1year. Relationships between various preoperative variables and the difference between predicted and actual weight loss were assessed using correlation, regression, and t-tests.RESULTS: There were 265 patients, with 176 age<18. The correlation coefficients for predicted and actual weight loss were 0.48 for patients age<18 and 0.38 for patients 18-21. On average, the proportion of predicted weight loss actually attained at 1year was 0.73. There were no statistically significant associations between calculator inaccuracy and patient age, sex, preoperative body mass index, or area deprivation index (all p>0.05).CONCLUSIONS: The MBASQIP calculator predictions show weak to moderate correlation with actual weight loss at 1year and should be used with caution when counseling pediatric patients considering MBS. This project underscores the importance of building multi-institutional collaborations and databases specific to the pediatric MBS context.

  View details for DOI 10.1007/s11695-025-08295-5

  View details for PubMedID 41068350

• Physics-Inspired Diffusion Probabilistic Models for Improved Denoising in Intracardiac Time Series. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference Ruiperez-Campillo, S., Rau, M., Ganesan, P., Brennan, K. A., Feng, R., Bandyopadhyay, S., Rogers, A. J., Narayan, S. M., Vogt, J. E. 2025; 2025: 1-5

  Abstract

  Intracardiac electrophysiological (EP) signals are frequently contaminated by diverse noise sources, posing a major obstacle to accurate arrhythmia diagnosis. We hypothesized that a physics-inspired conditional denoising diffusion probabilistic model (cDDPM) could outperform both classical filters and variational autoencoders by preserving subtle morphological features. Using 5706 monophasic action potentials from 42 patients, we introduced a range of simulated and real EP noise, then trained the cDDPM in an iterative process analogous to Brownian motion. The proposed model achieved superior performance across RMSE, PCC, and PSNR metrics, confirming its robustness against complex noise while maintaining essential signal fidelity. These findings suggest that diffusion-based methods can significantly enhance the clinical utility of EP signals for arrhythmia management and intervention.Clinical Relevance- We propose a denoising diffusion probabilistic model to reconstruct intracardiac signals in the presence of complex noise, which holds the potential to enhance diagnostic accuracy in EP procedures and inform more targeted treatment strategies.

  View details for DOI 10.1109/EMBC58623.2025.11252692

  View details for PubMedID 41336909

• Comparing Phenotypes for Acute and Long-Term Response to Atrial Fibrillation Ablation Using Machine Learning. Circulation. Arrhythmia and electrophysiology Ganesan, P., Pedron, M., Feng, R., Rogers, A. J., Deb, B., Chang, H. J., Ruiperez-Campillo, S., Srivastava, V., Brennan, K. A., Giles, W., Baykaner, T., Clopton, P., Wang, P. J., Schotten, U., Krummen, D. E., Narayan, S. M. 2025: e012860

  Abstract

  BACKGROUND: It is difficult to identify patients with atrial fibrillation (AF) most likely to respond to ablation. While any arrhythmia patient may recur after acutely successful ablation, AF is unusual in that patients may have long-term arrhythmia freedom despite a lack of acute success. We hypothesized that acute and chronic AF ablation outcomes may reflect distinct physiology and used machine learning of multimodal data to identify their phenotypes.METHODS: We studied 561 consecutive patients in the Stanford AF ablation registry (66±10 years, 28% women, 67% nonparoxysmal), from whom we extracted 72 data features of electrograms, electrocardiogram, cardiac structure, lifestyle, and clinical variables. We compared 6 machine learning models to predict acute and long-term end points after ablation and used Shapley explainability analysis to contrast phenotypes. We validated our results in an independent external population of n=77 patients with AF.RESULTS: The 1-year success rate was 69.5%, and the acute termination rate was 49.6%, which correlated poorly on a patient-by-patient basis (phi coefficient=0.08). The best model for acute termination (area under the curve=0.86, Random Forest) was more predictive than for long-term outcomes (area under the curve=0.67, logistic regression; P<0.001). Phenotypes for long-term success reflected clinical and lifestyle features, while phenotypes for AF termination reflected electrical features. The need for AF induction predicted both phenotypes. The external validation cohort showed similar results (area under the curve=0.81 and 0.64, respectively) with similar phenotypes.CONCLUSIONS: Long-term and acute responses to AF ablation reflect distinct clinical and electrical physiology, respectively. This de-linking of phenotypes raises the question of whether long-term success operates through factors such as attenuated AF progression. There remains an urgent need to develop procedural predictors of long-term AF ablation success.

  View details for DOI 10.1161/CIRCEP.124.012860

  View details for PubMedID 39925268

• Engineering of Generative Artificial Intelligence and Natural Language Processing Models to Accurately Identify Arrhythmia Recurrence. Circulation. Arrhythmia and electrophysiology Feng, R., Brennan, K. A., Azizi, Z., Goyal, J., Deb, B., Chang, H. J., Ganesan, P., Clopton, P., Pedron, M., Ruipe Rez-Campillo, S., Desai, Y., De Larochellière, H., Baykaner, T., Perez, M., Rodrigo, M., Rogers, A. J., Narayan, S. M. 2024: e013023

  Abstract

  Large language models (LLMs), such as ChatGPT, excel at interpreting unstructured data from public sources, yet are limited when responding to queries on private repositories, such as electronic health records (EHRs). We hypothesized that prompt engineering could enhance the accuracy of LLMs for interpreting EHR data without requiring domain knowledge, thus expanding their utility for patients and personalized diagnostics.We designed and systematically tested prompt engineering techniques to improve the ability of LLMs to interpret EHRs for nuanced diagnostic questions, referenced to a panel of medical experts. In 490 full-text EHR notes from 125 patients with prior life-threatening heart rhythm disorders, we asked GPT-4-turbo to identify recurrent arrhythmias distinct from prior events and tested 220 563 queries. To provide context, results were compared with rule-based natural language processing and BERT-based language models. Experiments were repeated for 2 additional LLMs.In an independent hold-out set of 389 notes, GPT-4-turbo had a balanced accuracy of 64.3%±4.7% out-of-the-box at baseline. This increased when asking GPT-4-turbo to provide a rationale for its answers, requiring a structured data output, and providing in-context exemplars, rose to a balanced accuracy of 91.4%±3.8% (P<0.05). This surpassed the traditional logic-based natural language processing and BERT-based models (P<0.05). Results were consistent for GPT-3.5-turbo and Jurassic-2 LLMs.The use of prompt engineering strategies enables LLMs to identify clinical end points from EHRs with an accuracy that surpassed natural language processing and approximated experts, yet without the need for expert knowledge. These approaches could be applied to LLM queries for other domains, to facilitate automated analysis of nuanced data sets with high accuracy by nonexperts.

  View details for DOI 10.1161/CIRCEP.124.013023

  View details for PubMedID 39676642

• AUTOMATED, ACCURATE IDENTIFICATION OF VENTRICULAR TACHYCARDIA FROM ELECTRONIC HEALTH RECORDS USING NATURAL LANGUAGE PROCESSING Brennan, K., Azizi, Z., Feng, R., Goyal, J., Liu, X., Ganesan, P., Ruiperez-Campillo, S., Baykaner, T., Badhwar, N., John, R. M., Viswanathan, M., Perino, A., Wang, P. J., Rogers, A. J., Narayan, S. M. ELSEVIER SCIENCE INC. 2024: 2644

  View details for Web of Science ID 001324901502678