Research

Yehuda is a Postdoctoral Associate at Rutgers New Jersey Medical School and a Consultant to Rutgers Cancer Institute. He develops fairness-aware deep learning and computer vision methods for health applications, with a focus on skin cancer detection and on Alzheimer’s disease and related dementias (ADRD). Across both areas, his goal is to build robust, clinically relevant AI systems that perform equitably across diverse patient populations.
Alzheimer’s disease and cognitive aging: At Rutgers New Jersey Medical School, Yehuda develops AI and deep learning models for cognitive aging and ADRD, with an emphasis on equitable performance in diverse populations. Working with Dr. Sharon Sanz-Simon and Prof. Michal Schnaider-Beeri and the research team, he designs predictive models / analyzes multimodal clinical and imaging data to support early detection and risk assessment.
Skin cancer detection: Yehuda builds and evaluates deep learning classifiers that distinguish benign from malignant skin lesions in clinical dermatology images. His research investigates data-centric subset and coreset selection strategies that improve model accuracy while reducing performance gaps across skin tone subgroups. He develops the full training and testing workflow behind reliable AI models, including how training data are selected, how models are tuned, and how performance is validated on held-out patients. Rather than reporting only overall accuracy, he examines whether a model performs equally well across skin tones by comparing key clinical outcomes such as cancer detection sensitivity (how often cancer is correctly identified) and false-alarm rates (how often benign cases are flagged as cancer). As a consultant to Rutgers Cancer Institute, he contributes AI and data science expertise to oncology research. He collaborates with dermatologists, oncologists, and radiologists to make sure his methods and evaluations match clinical needs and real-world use.
Methods and modalities. Yehuda’s work draws on computer vision, deep learning, and Convolutional Neural Networks (CNNs) as foundational tools for image understanding. It increasingly incorporates Vision-Language Models (VLMs) and Large Language Models (LLMs) for multimodal reasoning and language-driven analysis in clinical settings. His projects span a wide range of medical imaging modalities, including X-ray, CT, MRI, ultrasound, PET, and pathology/histopathology. This breadth reflects his commitment to methods that are technically robust, transferable across imaging contexts and clinical settings, and equitable across patient populations.






↑Back to Top