Cardiovascular Analytic Intelligence Initiative
Cardiovascular Analytic Intelligence Initiative (CV-Ai2) aims to find new and innovative ways to better utilize the data generated through clinical activity to design solutions that help solve important clinical problems. These solutions are then deployed in the hospital setting and directly evaluated at the physician and patient level.
In This Section:
From Analytics to Analytic Intelligence
Among the hundreds of predictive models developed for cardiovascular disease, only <0.1% actually end up routinely used in clinical practice. The Cardiovascular Analytic Intelligence Initiative (CV-Ai2) was created at Johns Hopkins University to specifically address this translational crisis. The key to address the current limitations of predictive models will be to move beyond data analytics and artificial intelligence and focus on analytic intelligence. Analytic intelligence is a multi-step process in which predictive models are created while considering the full requirements for their deployment and evaluation, which is an intrinsic part of the process. The move to analytic intelligence will be necessary to fully realize the potential of machine learning and artificial intelligence in clinical settings.
Analytics to Intelligence
We develop the infrastructure necessary for real-time acquisition and utilization of clinical data in predictive models. This includes developing the technology for real-time data extraction and processing and integration of predictive models into the clinical IT infrastructure.
We develop and test new strategies to create predictive models for clinical use using best-in-breed biostatistical methods and machine learning. Our focus is on methods that allow for the utilization of complex and deep data without using data reduction methods.
- Shelby Kutty, M.D., Ph.D., Chair
- Cedric Manlhiot, Ph.D., Director
- Ari Cedars, M.D., Member
- Benjamin Barnes, M.D., Member
- Lasya Gaur, M.D., Member
- Vivek Jani, M.S., Medical Student
- Joseph Shin, B.S., Medical Student
- Junzhen Zhan, M.D., Ph.D., Research Fellow
Heart Life Trajectory Project
Patients with structure heart disease often progress toward heart failure over time. The Heart Life Trajectory project aims to consolidate lifelong data on thousands of patients with structural heart disease, and use data analytics to predict progression of heart function and the optimal timing and modalities for therapeutic or surgical interventions.
Cardiac Outpatient Warning System
Physicians often only get a snapshot of the status of patients with stable heart disease when they see them in outpatient settings. Unfortunately, the clinical status of many patients deteriorates between visits, and physicians might miss the narrow window of opportunity to intervene. The Cardiac Outpatient Warning System uses predictive and real-time analytics of data from wearables to detect subtle changes in heart function to alert both the patient and the care provider it might be time for evaluation.
Predictive Decision Support Tools
Decision support tools are ubiquitous in medicine, but their use of data is often limited both in scope and complexity. The CV-Ai2 team works on numerous projects where complex predictive algorithms are developed, implemented in real-time clinical decision support, and evaluated. These tools will allow for individualization of therapy, better matching of patients to therapy, and hopefully better patient outcomes.
Statistical Learning for Risk Predictive Models
The classic way of creating predictive models uses probabilistic methods. They work very well in a narrow set of circumstances (common risk factors) and often rely on data reduction techniques where a large proportion of the available data is not used. At CV-Ai2, we use statistical learning methods to avoid the need for data reduction, so they can perform well in a wider set of circumstances.
We offer a variety of opportunities for students in medicine, public health, data sciences and computer sciences to contribute to research projects. There are many ways to contribute to projects and previous coding or data analytics experience is not necessary for all projects. The only prerequisite is previous completion of a biostatistics course at the undergraduate level.
Students typically dedicate at least 4–6 hours per week on projects with flexible schedule. Students who provide substantial contributions on a project will be invited to lead manuscript preparation and presentation of results at national and international conferences.
Positions for graduate students and post-doctoral fellows will be available starting the second half of 2020. Details to come.
For CV-Ai2’s vision to become a reality, our academic, clinical and industry partners must come together and work toward a common goal. CV-Ai2 is grateful for the partners who help make our vision possible, and we are always looking for more collaborators. If you share our vision and want to get involved, please do not hesitate to contact us at firstname.lastname@example.org.