Michael Grasso - Research

Research Interests

The focus on my research is on innovative applications of biomedical informatics that expand the scope of clinical medicine, have a strong theoretical basis in computer science, and are of strategic importance to the University of Maryland Medical Center. My research interests include clinical decision support, clinical data mining, ubiquitous computing, database engineering, and software engineering, which I am applying to chronic disease management, patient safety and quality improvement, and mental health disorders.

Current Research Projects

  • Knowledge Representation and Reasoning with Big Data - We are developing new methods for knowledge representation and reasoning that are optimized for very large clinical repositories, and which can be correlated with genomic and environmental data. Our clinical focus is on coronary artery disease, diabetes, lower respiratory disease, rheumatoid arthritis, certain cancers, chronic pain, addiction, and mental illness. This work will lead to new approaches to quantify clinical risk and enable clinical decision support. It can be applied to disease prediction, critical event prediction, and treatment efficacy. We are currently working with a large clinical repository with data on more than 20 million patients from roughly 150 medical centers and 800 outpatient clinics that contains over one billion patient encounters with associated vital sign, laboratory, radiology, health screening, and pharmacy data.

  • Patient Safety and Quality Measures in Emergency Medicine - Patient safety is one of the nation's most important health care challenges. The widely-cited report by the Institute of Medicine estimates that as many as 44,000 to 98,000 people die in U.S. hospitals each year as the result of lapses in patient safety. This is especially important in the emergency department, where health care teams are challenged to rapidly diagnose and treat multiple patients, some of whom present with potentially life-threatening illness. Safety events can be difficult to measure, in part, because identification often depends on self-reporting, and also because we tend to discover only those events that result in significant harm to patients. We are therefore developing new computational methods to identify patient safety events that are based on semantic analysis and situational awareness, and which can be used by health care teams for computationally-enabled clinical decision support. We are also conducting research in resource utilitzation and recidivism in emergency medicine, with a focus on co-morbidities, key risk factors, utilization patterns, and clinical workflow.

  • Computer-Aided Risk Assessment for Suicide, Mental Illness, and Addiction - Suicide is a significant and preventable public health problem, and the tenth leading cause of death in the U.S. In addition, war veterans have a higher risk of suicide, due to a combination of war-related mental illness and physical health problems. While certain known risk factors exist, such as mental illness and addiction, suicidal behavior is complex and often hard to predict. We are therefore developing new computation approaches for risk assessment for suicide surveillance, mental illness, and addiction. This includes the development of a risk ontology along with a set of risk indicators that will be validated through an analysis of clinical, environmental, and genomic data.

Graduate Students

  • Miguel Ossandon, 2009, PhD Student, visual sensory substitution
  • Xianshu Zhu, 2009, MS Student, clinical image processing
  • Dashana Dalvi, 2009-2010, MS Student, clinical image processing
  • David Chapman, 2010, PhD Student, clinical image processing with semantic metadata in cell biology
  • Justin martineua, 2010-2011, PhD Student, clinical image processing with semantic metadata in cell biology
  • Ronil Mokashi, 2009-2011, MS Student, clinical image processing with semantic metadata in cell biology
  • Ashwini Lahane, 2010-2011, MS Student, Thesis Committee, clinical image processing
  • Ronil Mokashi, 2011, MS Student, Thesis Committee, clinical image processing
  • Dashana Dalvi, 2011, MS Student, Thesis Committee, genomic prediction of type 2 diabetes
  • Aniket Bochare, 2011-2012, MS Student, Thesis Committee, domain knowledge and genomic prediction
  • Ronil Mokashi, 2011-2012, MS Student, Thesis Committee, genomic association and machine learning
  • Zachary Kurtz, 2012, MS Student, Thesis Committee, simultaneous feature and cost estimation
  • Eduardo Llamas, 2012, MS Student, Thesis Committee, extraction of drug-gene relationships
  • Soma Das, 2011-2012, MS Student, Thesis Advisor, genomic prediction of coronary artery disease
  • Matthew Gately, 2011-2012, MS Student, Thesis Advisor, personalized medicine
  • Isaac Mativo, 2012, PhD Student, Thesis Advisor, personalized medicine
  • Sandeep Regmi, 2014, Medical Resident, wearable computing for clinical decision support

Past Research Projects

  • Syndromic Surveillance

Syndromic surveillance is a process used for the early detection and monitoring of disease outbreaks attributable to biologic terrorism or natural causes. We explored new approachs to syndromic surveillance by applying machine learning algorithms to both clinical data and social media.

  • A Computerized Approach to Glycemic Control

Lessons learned from prior efforts were used to design an optimal approach to computerize insulin protocols for critical care and general medicine patients that better fits into the existing physician workflow.

  • Identification and Analysis of a TSH-Associated SNP

An analysis of a TSH-associated SNP was performed to help identify which gene/variant was the most likely functional unit.

  • Ventricular Assist Devices and Gastrointestinal Bleeding

A retrospective analysis of consecutive VAD recipients found that age was the only independent predictor of gastrointestinal bleeding, and that nonpulsatile VADs were not associated with an increase in gastrointestinal bleeding versus there pulsatile counterparts.

  • Efforts to Minimize Adverse Events by Cross-Covering Physicians

An analysis of process improvement efforts regarding on the transfer or care was performed, and a software framework was developed to increase the accuracy and safety of this process.

  • Medical Response to Terrorist Attacks and Other Disasters

A redundant communication framework was developed to mobilize medical personnel and transfer information during a medical disaster.

  • Clinical Trials Data Collection Using Palm Computers

A handheld data collection system was developed to update and validate clinical information obtained at the bedside or in the field.

  • Voice-Driven Pathology Data Management System

A voice-driven system was developed to collect clinical data in hands-busy and eyes-busy environments.

  • Computational Image Classification

We developed image classification techniques to extract and map image features to biological characteristics. We applied this algorithm to laparoscopic surgery videos, in order to segment the videos into clinical steps and to recognize potentially unsafe actions. We also used this algorithm to extract geometric image feature from smooth muscle images to help characterize cell-to-matrix interactions.

  • Intelligent Software Agents for Disaster Management

We developed an agent framework for the Department of Homeland Security to support a peer-to-peer network of disaster responders using the Unified Incident Command and Decision Support (UICDS) platform. This is a middleware foundation that enables commercial and government incident management technologies to share information and support decisions for the National Response Framework and National Incident Management to prevent, protect, respond, and recover from natural, technological, and terrorist events.

Fisher Exact Test  Odds Ratio  Logistic Regression  T Test  CI  Pearson  Other  ANOVA  Sample Size

NCBI  HapMap  Blast  David  Seattle SNP  GenMAPP  Vista  Ingenuity