Harvard Catalyst Profiles

The Cardiovascular Imaging Research Center of Mass General Hospital and Harvard Medical School is seeking outstanding MD or PhD doctoral students for a remote or in-person project to build deep learning models to predict cardiovascular disease (e.g., atherosclerosis, heart attack, stroke) risk from routine medical imaging (chest CT, retinal fundoscopy, coronary CT). Our lab has high quality datasets with tens of thousands of individuals with imaging and adjudicated outcomes (PROMISE Douglas PS et al. NEJM 2015; Framingham Heart Study Hoffmann U et al. JAMA Cardiology 2017; REPRIEVE Hoffmann U, Lu MT et al Am Heart J 2019; NLST Berg CD et al NEJM 2011; Jackson Heart Study; and the Mass General Brigham Biobank). Prerequisite: Coding experience, especially in Python with a commitment to learn PyTorch or Tensorflow deep learning frameworks. Some recent representative publications: JACC Cardiovascular Imaging: https://www.sciencedirect.com/science/article/pii/S1936878X21000681?via%3Dihub Annals of Thoracic Surgery: https://www.sciencedirect.com/science/article/pii/S0003497522007226?via%3Dihub Nature Communications: https://www.nature.com/articles/s41467-021-20966-2

The Cardiovascular Imaging Research Center of Mass General Hospital and Harvard Medical School is seeking outstanding MD or PhD doctoral students for a remote or in-person project to develop a large language model (LLM)-based pipeline for patient phenotyping (e.g., based on clinic notes and discharge summaries, is it likely that this patient had a heart attack?) to test whether such systems can improve phenotyping of Mass General Brigham patients beyond International Classification of Disease (ICD) codes. This study will be conducted using high-quality datasets from Mass General Brigham with gold-standard outcomes based on previously performed manual chart review of thousands of patients. Prerequisite: Coding experience, especially in Python. Prior Natural Language Processing or LLM prompting experience is a plus, but not required.

The Cardiovascular Imaging Research Center of Mass General Hospital and Harvard Medical School is seeking outstanding MD or PhD doctoral students for a remote or in-person project to study how deep learning applied to routine imaging can improve understanding of the genetic architecture of disease (What genes “prime” a person to have this imaging-based phenotype?). Our lab has several high-quality datasets with tens of thousands of individuals with imaging and associated genotype data (the UK Biobank, the Mass General Brigham Biobank, the Framingham Heart Study, and the Jackson Heart Study). Prerequisite: Some experience with statistical genetics, especially genome-wide association studies, experience with AI or deep learning is not required, but is a plus. A recent representative publication related to this project in Circulation: https://www.ahajournals.org/doi/10.1161/CIRCULATIONAHA.121.057709