Fundamental Intelligent Building Blocks of the Intensive Care Unit (ICU) of the Future: Intelligent ICU of the Future

Delirium, Pain

The objective of this project is to create deep learning and machine learning models capable of recognizing patient visual cues, including facial expressions such as pain and functional activity. Many important details related to the visual assessment of patients, such as facial expressions like pain, head and extremity movements, posture, and mobility are captured sporadically by overburdened nurses or are not captured at all. Consequently, these important visual cues, although associated with critical indices, such as physical functioning, pain, and impending clinical deterioration, often cannot be incorporated into clinical status. The study team will develop a sensing system to recognize facial and body movements as patient visual cues. As part of a secondary evaluation method the study team will assess the models ability to detect delirium.

Pain is a critical national health problem with nearly 50% of critical care patients experience significant pain in the Intensive Care Unit (ICU). The under-assessment of pain response is one of the primary barriers to the adequate treatment of pain in critically ill patients, associated with many negative outcomes such as chronic pain after discharge, prolonged mechanical ventilation, longer ICU stay, and increased mortality risk. Nonetheless, many ICU patients are unable to self-report pain intensity due to clinical conditions, ventilation devices, and altered consciousness. Currently, behavioral pain scales are used to assess pain in nonverbal patients. Unfortunately, these scales require repetitive manual administration by overburdened nurses. Moreover, prior work suggests that nurses caring for quasi-sedated patients in critical care settings have considerable variability in pain intensity ratings. Furthermore, manual pain assessment tools lack the capability to monitor pain continuously and autonomously. Together, these challenges point to a critical need for developing objective and autonomous pain recognition systems. Delirium is another common complication of hospitalization that poses significant health problems in hospitalized patients. It is most prevalent in surgical ICU patients with diagnosis rates up to 80%. It is characterized by changes in cognition, activity level, consciousness, and alertness. Delirium typically leads to changes in activity level and alertness that pose additional health risks including risk of fall, inadequate mobilization, disturbed sleep, inadequate pain control, and negative emotions. All of these effects are difficult to monitor in real-time and further contribute to worsening of patient's cognitive abilities, inhibit recovery, and slow down the rehabilitation process. Though about a third of delirium cases can benefit from intervention, detecting and predicting delirium is still very limited in practice. Current Delirium assessments need to be performed by trained healthcare staff, are time consuming, and resource intensive. Due to the resources necessary to complete the assessment, delirium is often assessed twice per day, despite the transient nature of the disease state which can come and go undetected between the assessments. Jointly these obstacles demonstrate a dire need for real-time autonomous delirium detection. The investigators hypothesize that the proposed model would be able to leverage accelerometer, electromyographic, and video data for the purpose of autonomously quantifying patient facial expressions such as pain, characterizing functional activities, and delirium status. Rationalizing that autonomous visual cue quantification and delirium detection can reduce nurse workload and can enable real-time pain and delirium monitoring. Early detection of delirium offers patients the best chance for good delirium treatment outcomes.

No related papers found yet.

United States