In the US, more than 5.7 million adults would be admitted into the intensive care units (ICU) annually. Although a substantial amount of data will be captured from each of them, the granular indications such as the impact of background noises and lighting on patients and aspects that reflect a person’s well-being like facial expressions, functional status, mobility, and so on are seldom measured. Manually accounting for these details may lack objectivity and also increases the workload of staff.

An intelligent ICU for more timely and targeted interventions

As such, a group of researchers at the University of Florida tried tapping onto artificial intelligence (AI) and pervasive sensing (i.e., use of sensors to get hold of surrounding information) to monitor patients in an autonomous and continuous manner. They hope the method will garner physicians additional, real-time knowledge of patients to develop more timely and targeted life-saving interventions.

To do so, the research team tried to instill intelligence into a traditional ICU by making use of three wearable accelerometer sensors, a light sensor, a sound sensor, and a high-resolution camera to pick up details from ICU patients and their environment. Computer vision and deep learning would then be deployed to analyze patients’ face, posture, facial expressions, facial action units and head pose from the videos and data collected from wearable accelerometer sensors worn on the wrist, ankle and arm.

In addition, researchers also assessed the effects of light intensity and sound pressure on patients’ sleep quality using the Freedman Sleep Questionnaire and the frequency of visits made by visitors and medical staff into the ICU. All these were backed up by patients’ clinical records obtained from the electronic health record (EHRs) systems.

22 patients with and without ICU delirium were recruited in the pilot study to determine whether pervasive sensing and AI can be used to characterize the impacts of the patients’ environment and their functional status and pain. Researchers found that pervasive sensing and AI are capable of autonomous monitoring of patients in ICU setting.

Pervasive sensing and AI for autonomous monitoring of ICU patients

The cost of building one system with sensors and cameras is estimated to be around $300 US dollar, which is relatively low as compared to the thousands of dollars currently spent on each ICU patient per day. Besides, devices can also be reused for other patients, further cutting down on expenditure and potential wastage. On the other hand, the collected data revealed that ICU does have a significant impact on delirious patients’ circadian rhythm, a finding which coincided with literature.

Overall, delirious patients reported a lower capability to fall asleep in the ICU and a higher degree of getting disturbed by the continuous lighting as compared to non-delirious patients. Delirious patients are also more affected by visitor disruptions while non-delirious patients experienced more disruptions in the day. Researchers accounted the differences as a result of interactions with others. Nevertheless, there was no significant difference in delirious patients’ perception of noise as compared to non-delirious patients.

Data from the wearable accelerometer sensors showed that patients’ activity in the wrist was significantly different between delirious and non-delirious patients but this was not the case for arm and ankle. Researchers believed the insight could potentially lead to a reduction of on-body sensors to monitor certain conditions. As this is only a pilot study, researchers are unsure if AI and pervasive sensing will remain unchallenged in a larger and more diverse pool of ICU patients.

Furthermore, the study did not consider of the medications that these ICU patients were having, which also affect a patient’s well-being. Researchers also failed to regard privacy and possible loss of footages or signals due to the multitude of medical devices found on an ICU patient and the number of staffs entering and exiting the premises. They hope that future studies will address these challenges. All related findings can be found in Scientific Reports.

Join us on 22 September at the AIMed ICU virtual event for an informative overview of how AI, robotics, virtual/augmented/mixed realities, and many others that are influencing the realm of ICU and strategies to deploy these technologies in the clinical setting. Register your interest or get a copy of the agenda here today!


Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.