Our team is broadly interested in developing computational tools and machine learning algorithms for analysis of
medical data, with the goal of phenotyping patients, tracking patients' trajectory, predicting impending events, and
optimizing treatment strategies to avoid/mitigate adverse outcomes.
To accomplish these, we bring together concepts and tools from signal processing, information theory, control theory,
optimization, and machine learning to design physiologically inspired models and predictive analytic and sequential
decision-making algorithms. Additionally, we are involved in a number of translational and regulatory activities,
which include close collaborations with health IT professionals, Quality Management System (QMS) engineers,
FDA regulatory experts, bedside clinicians, and hospital quality improvement and patient experience teams.
Together these allow us to close the loop on building functional and sustainable healthcare Apps and provide
the much-needed evidence to support the deployment of data-driven clinical and patient-facing decision support systems.
Application areas include critical care, computational neuroscience and psychiatry, and wearable technology.
See our journal club, Deep Learning Course (Spring 2017) , and blog posts.
A major focus of our research has been in the intensive care unit (ICU) where we have been
developing advanced machine learning algorithms capable of meaningfully summarizing large volumes of continuously
measured patient data (e.g., vital signs time series, lab values, medication records, clinical notes, etc.),
with the goal of timely prediction of potentially life-threatening clinical events (e.g., Sepsis).
(Find out more)
We also have a number of active collaborations at the intersection of machine learning and
neuroscience/psychiatry. We believe computational and quantitative approaches to phenotyping
clinical states of patients (e.g., during treatment of depression), can prove useful for both
monitoring outcomes and justifying treatment interventions. (Find out more)
Our team is also actively involved in extending our monitoring and predictive algorithms beyond the
inpatient setting, using ambulatory and wearable technologies,
where handling of noise and recording artifacts requires novel signal processing techniques.
(Find out more)