Harvard Catalyst Profiles

In this project, student will leverage computational, statistical, and artificial intelligence (AI) methods to harness the potential of "big data" across a variety of domains to advance research and enhance clinical care in neurology. Student will apply AI to a rich spectrum of data types—wearables, sensors, brain imaging, and electronic health records—in order to drive forward both research and clinical decision-making Some example projects include the use of large language models (LLMs), such as GPT, for the analysis of unstructured text in electronic health records. By extracting and interpreting complex clinical narratives, the student will aim to uncover insights into disease patterns, treatment outcomes, and patient trajectories. This approach not only promises to support research studies, but also aims to personalize and optimize patient care management. Alternatively, the student may explore the potential of smartphone and wearable technologies in the development of digital biomarkers. By analyzing voice and time-series data collected from these devices, the project seeks to identify and validate new biomarkers that can detect and monitor the progression of neurological diseases. This project has the potential to enable continuous patient monitoring, early detection, and intervention strategies. Overall, student will use a multidisciplinary approach that harnesses the power of big data and AI for both research and clinical care.

Traditional methods of neuropathological assessment are manual and labor-intensive. With the integration of AI, particularly foundation models, there is a significant opportunity to develop automated techniques that could streamline and enhance the efficiency and accuracy of these assessments. Currently, most foundation models in digital pathology have been developed predominantly with datasets derived from cancer patient samples, resulting in a scarcity of models trained on brain tissue data. This project aims to bridge this gap by having students evaluate the effectiveness of foundational AI models for differential diagnosis using whole slide images from brain autopsies. The students will assess these models’ capabilities in diagnosing and understanding various neuropathological conditions, including determining Braak stage, identifying hippocampal sclerosis, and evaluating vascular injuries.