The lab is affiliated with Mayo Clinic, Arizona and SCAI, Arizona State University. Our members are computer scientists and clinical professionals with passion for AI in healthcare.
Welcome to the BanerjeeLab (A3I Hub, Mayo Clinic)
At BanerjeeLab, we are pioneering the fusion of unstructured medical data—including radiology and pathology images, 3D anatomical models, and clinical notes—to develop advanced AI models that support clinical diagnosis, treatment planning, and precision medicine.
🎯 Our Focus
We are particularly interested in the longitudinal inference of patient data—understanding disease progression over time—and building pathways to deploy AI tools at the point-of-care. Our goal is to empower clinicians with real-time, data-driven insights for prevention, screening, and treatment of cancer and cardiovascular diseases.
🔬 Core Research Areas
AI Fairness in Healthcare
Multimodal Data Fusion
Clinical translation of AI models
Anomaly Detection
Large Language Models (LLMs)
Vision-Language Models (VLMs)
MedAI Journal club every week – joint with Stanford University, Emory University and NIH.
Collaborations: We have several multi-institutional research projects in collaboration with multiple academic centers and industries in US (including Stanford, Emory, MIT, Duke, SUNY Buffalo, Moffitt Cancer Center, UCSF, Philips healthcare, Carl Zeiss, Genetech, Arterys Inc) and Europe (CNR, Université Paris Descartes, HES-SO Valais) related to NLP and prediction. The application area of the developed methodologies spans from breast cancer, liver cancer, prostate cancer up to cardiovascular diseases, sepsis.
Competencies: Natural Language Processing, Computer Vision, Medical image analysis, deep learning, Multimodal data, Predictive modeling.
Open graduate student support and postdoc position in both Computer vision and NLP: please send CV and a letter of research interest to imon.banerjee[at]asu.edu
News
New NCI ITCR R21: Prostate Cancer
I’m excited to share that our team at #MayoClinic Arizona has received a new R21 award from the National Cancer Institute (#NCI) to advance our work in AI-enabled patient communication and support for prostate cancer.
BTW, this is Dave Le first grant as PI!!
A major focus of this grant will be developing and evaluating an agentic AI solution that can leverage clinical data and patient context to provide personalized, timely, and meaningful communication and support throughout the prostate cancer care journey.
For patients with prostate cancer, navigating care can involve complex decisions, evolving treatment plans, symptoms and side effects, and many questions that arise outside of clinical visits. We believe there is an important opportunity for agentic AI to go beyond traditional, static patient-facing tools—reasoning over relevant context, proactively supporting communication, and helping patients better navigate their care while remaining connected to their clinical team.
Our goal is to investigate how these emerging AI capabilities can be developed safely, responsibly, and in alignment with clinical workflows to augment—not replace—the human relationships at the center of cancer care.
I’m grateful to our collaborators, colleagues, and mentors whose contributions and support made this work possible.
We’re excited to take the next step toward a new generation of intelligent, patient-centered support for prostate cancer care.
#ProstateCancer #CancerResearch #AgenticAI #ArtificialIntelligence #DigitalHealth #PatientSupport #PrecisionMedicine #NCI #MayoClinic
A new editorial in Circulation
Great to see our work generating discussion beyond the paper itself.
A new editorial in Circulation: Cardiovascular Imaging offers a thoughtful perspective on our study, “Beyond Doppler: Scalable AI Detection of LVOT Obstruction in HCM,” and what AI might contribute to the evaluation of hypertrophic cardiomyopathy.
Editorial
One of the most interesting takeaways: AI is not replacing Doppler. Instead, routine resting 2D echocardiographic videos may contain structural and spatiotemporal signals that reveal a predisposition to LVOT obstruction—even when obstruction only becomes apparent with provocation.
That distinction is important. The potential role of AI here is not to replace careful Doppler interrogation, physiologic provocation, or clinical judgment. It is to serve as an additional safeguard—helping identify patients whose resting examination may look reassuring but who may warrant more intensive provocative evaluation.
The editorial also appropriately highlights the work still ahead, including broader external validation, prospective evaluation, and understanding performance across laboratories and provocative protocols.
Grateful to Daniele Massera and Mark V. Sherrid for their thoughtful perspective on our work—and to the entire team behind this study.
Reza Arsanjani Owen Crystal, M.S. Steven Lester MD Chadi Ayoub
Paper: Crystal O, Farina J, Scalia I, et al. Beyond Doppler: Scalable AI Detection of LVOT Obstruction in HCM. Circulation: Cardiovascular Imaging. 2026. https://www.ahajournals.org/doi/full/10.1161/CIRCIMAGING.126.020318
#ArtificialIntelligence #Echocardiography #HypertrophicCardiomyopathy #CardiovascularImaging #MedicalAI
Mayo News about our publication
Hypertrophic cardiomyopathy can be difficult to diagnose. New research aims to help clinicians better identify high-risk patients, supporting earlier diagnosis, earlier treatment and better outcomes.
Watch to learn how this work could make a difference for patients.
Read more about how a Mayo Clinic AI model helps clinicians detect heart obstruction using routine ultrasound images: