Marcelo Di Carli, Scott Solomon, Ron Kikinis, Sharmila Dorbala, Rahul Deo
The growing role and new approaches in image analysis, data science and outcomes research has fostered major expansion in research in these areas in our Program. The establishment of the MGH & BWH Center for Clinical Data Science (CCDS), a multidisciplinary center, involving machine learning scientists, data engineers, software engineers, innovation fellows, and faculty from MGH, BWH and Harvard Medical School with the aim of advancing clinical research, deep phenotyping and precision medicine, is emblematic of the overall importance of the research opportunities in this research area. Many of our faculty have strong and close ties with the CCDS and are developing collaborative efforts for many years. Fellows working in these areas will be trained in quantitative image analysis, machine learning, artificial intelligence, non-invasive risk assessment, comparative treatment studies and outcomes research.
Scott Solomon (Clinical Profile, Research Profile) and his group have been utilizing unsupervised machine learning approaches to phenogroup patients based on patterns of ventricular contraction. These approaches have been used to identify patients who are more likely to respond to cardiac resynchronization therapy (CRT) and to generate multidimensional phenotype maps in a broad range of patients with and at risk for heart failure. These approaches complement traditional statistical modeling techniques and are allowing for more comprehensive utilization of imaging information beyond standard measurements of ventricular size and function.
Marcelo Di Carli (Clinical Profile, Research Profile) and Sharmila Dorbala (Clinical Profile, Research Profile) are cardiologists with ongoing collaborations with external faculty that use large multicenter imaging registries to develop methods for fully automated analysis of nuclear cardiology data using novel algorithms and machine learning techniques. They have recently shown that deep learning improves automatic interpretation of myocardial perfusion images as compared to conventional quantitative methods.
Ron Kikinis (Clinical Profile, Research Profile) is a radiologist and the founding Director of the Surgical Planning Laboratory at BWH. He pioneered image processing algorithms and their use for extracting relevant information from medical imaging data and is the Principal Investigator of 3D Slicer, a free open source software platform for image analysis and visualization. His research activities include the development of novel tools for image analysis including segmentation, registration, visualization, and high-performance computing including machine learning and artificial intelligence.
Rahul Deo (Clinical Profile, Research Profile) is a cardiologist with expertise in human genetics, computational biology, and data science approaches to understanding mechanisms of cardiovascular disease. He joined the cardiovascular division at BWH in May 2018 and currently serves as the Chief Data Scientist at One Brave Idea. His research group has developed innovative applications of machine learning to patient phenotyping to bridge the gap between patient data and experimental models of heart disease.