Intelligent Intensive Care Unit
In the intensive care unit (ICU), time is precious in terms of patient condition, clinician attention, and money. Annually in the United States, 5.7 million patients are admitted to ICUs, with costs exceeding $82 billion.
Healthcare providers must monitor and assess patient acuity as a part of ICU care, but their assessments may be sporadic due to overburdened clinicians and only reflect static, imprecise moments of a patient’s condition due to manual calculations. As a result, there is a critical unmet need for real-time interpretable, dynamic, autonomous, and precise acuity prediction in the ICU that integrates visual assessment of the patient with continuous physiologic measurement and clinical data.
Therefore, the Intelligent ICU project (I2CU) aims to develop and validate real-time autonomous acuity assessment and prediction tools using novel deep learning (DL) and sensing technologies. Our hypothesis is that through integrating clinical and physiological data with real-time autonomous assessment of visual cues (e.g., pain, emotional distress, physical function), DL models will outperform existing clinical models for predicting patient acuity.
To create this tool, researchers are pursuing the following goals.
For Interpretable Acuity Prediction
Develop and validate an interpretable DL model to determine if it the DL model more accurate than current models in predicting transitions in the acuity of a patient’s daily clinical status, while providing interpretable information to physicians. Integrated electronic health record and continuous physiological data will help train and validate the DL model.
For Autonomous Visual Assessment
Develop a pervasive sensing system that autonomously captures 33 visual cues of critically ill patients and determine if it can provide accurate visual assessments compared to human experts.
For Real-Time Intelligence
Implement and evaluate an intelligent platform for real-time integration of autonomous visual assessment and acuity prediction in clinical workflows to determine accuracy in real-time prospective evaluation and to determine physicians’ satisfaction with the risk assessment process.
Successful application of the proposed technology would augment clinical decision-making in the fast-paced environment of ICU to promote earlier and more targeted interventions. Real-time, precise predictions of clinical trajectory reduce the need for manual, repetitive ICU assessments and capture patient aspects that were previously unnoticed. Ultimately, the results are expected to improve patient outcomes, decrease hospitalization costs, and reduce lifelong complications.
Select I2CU Publications
- “Potentials and Challenges of Pervasive Sensing in the Intensive Care Unit,” Davoudi A, Shickel B, Tighe PJ, Bihorac A, Rashidi P. Frontiers in Digital Health. 2022;4:773387. Published 2022 May 17.
- “Multi-Task Prediction of Clinical Outcomes in the Intensive Care Unit Using Flexible Multimodal Transformers,” Shickel B, Tighe PJ, Bihorac A, Rashidi P. arXiv. arXiv:2111.05431v1. Published 2021 Nov 9.
News Segments about I2CU
- “High-Tech Hospital Uses Artificial Intelligence in Patient Care,” NBC Nightly News, February 4, 2023.
- “Your Health: AI and the ICU,” News 10 WILX (NBC), February 18, 2022.
More I2CU Information
- Intelligent Intensive Care Unit (I2CU): Pervasive Sensing and Artificial Intelligence for Augmented Clinical Decision-Making, National Institutes of Health (NIH) RePORTER