Toyota AI seminar: Temporal Pattern Recognition in the ICU
Add to Google Calendar
Intensive Care Unit (ICU) patients are physiologically fragile and require vigilant monitoring and support. The myriad of data gathered from biosensors and clinical information systems has created a challenge for clinicians to assimilate and interpret massive volume of data. Physiologic measurements in the ICU are inherently noisy, multidimensional, and can readily fluctuate in response to therapeutic interventions as well as evolving pathophysiologic states. ICU patient monitoring systems may potentially improve the efficiency, accuracy and timeliness of clinical decision-making in intensive care. However, the aforementioned characteristics of ICU data can pose a significant signal processing and pattern recognition challenge—often leading to false and clinically irrelevant alarms. We have developed a temporal database of several thousand ICU patient records to facilitate research in advanced monitoring systems. The MIMIC-II database includes high-resolution physiologic waveforms such as ECG, blood pressures waveforms, vital sign trends, laboratory data, fluid balance, therapy profiles, and clinical progress notes over each patient's ICU stay. We quantitatively and qualitatively characterize the MIMIC-II database and include examples of clinical studies that can be supported by its unique attributes. A brief summary is provided of some challenges at the interface of Artificial Intelligence (AI) and ICU monitoring that can be explored with the MIMIC-II database. We also introduce a novel algorithm for identifying "similar" temporal patterns that may illuminate hidden information in physiologic time series. A novel temporal similarity metric is described based on a wavelet decomposition to characterize time series dynamics at multiple time scales. The compressed wavelet representation allows for the utilization of classical information retrieval algorithms based on a vector-space model. We demonstrate that statistical similarities between different patient time series may have meaningful physiologic interpretations in the detection of impending hemodynamic deterioration. As a generalized time series similarity metric, the algorithms that are described have applications in several other domains as well.
Mohammed Saeed is currently a fellow in Cardiovascular Medicine at the University of Michigan Cardiovascular Center and a trainee in the Physician-Scientist Program of the Department of Internal Medicine. He received an MD-PhD from MIT’s Department of Electrical Engineering and Computer Science and the Health Sciences and Technology (HST) Division of Harvard Medical School in 2008. His research was conducted in the Laboratory for Computational Physiology at MIT, and focused on medical database development and multi-scale physiologic time series analysis and pattern recognition. In 2006, his research was recognized at the American Medical Informatics Association (AMIA) annual conference with the “Most Distinguished Paper” Award. He has also been employed as a research scientist at Philips Healthcare in Andover, Massachusetts and contributed to the development of clinical decision support algorithms in ICU monitors and clinical information systems.