Researchers at Mount Sinai trained a deep-learning algorithm using video footage from the NICU to track infants’ movements and gather crucial neurological data, enabling precise monitoring of their actions.
Researchers have unveiled breakthrough findings from a novel AI-based tool, potentially paving the way for a minimally invasive and scalable method of continuous neurological monitoring in neonatal intensive care units (NICUs). This game-changing technology offers real-time, unprecedented insights into infant health.
Each year, more than 300,000 newborns are admitted to neonatal intensive care units (NICUs) across America. The assessment of toddler alertness is considered the most sensitive aspect of the neurologic examination, providing crucial insight into the overall integrity of the central nervous system throughout development. Unforeseen neurologic decline in neonatal intensive care units (NICUs) can have catastrophic consequences. Despite advances in cardiorespiratory telemetry, which continuously monitors infants’ heart and lung function in NICUs, neurotelemetry remains largely absent from the majority of NICUs, despite years of research in electroencephalography (EEG) and specialized neuro-NICU development. Neurological status is assessed periodically through physical examinations that are inherently imperfect and prone to missing subtle, subacute changes.
Researchers at Mount Sinai posited that employing a computer vision-based approach to track infant movements in the neonatal intensive care unit (NICU) could potentially forecast neurological changes.
The innovative “Pose AI” technology utilizes machine learning to track anatomical markers from video data, having a profound impact on both athletics and robotics.
Researchers at The Mount Sinai Hospital trained an artificial intelligence algorithm using over 16 million seconds of video footage from a diverse group of 115 infants in the NICU, leveraging their existing protocol for continuous video electroencephalography (EEG) monitoring. Researchers have successfully showcased Pose AI’s ability to accurately track toddler milestones from visual data. Utilizing anatomical references from the video feed, researchers were able to accurately predict two critical scenarios: sedation and cerebral impairment.
“Despite widespread adoption of video cameras in neonatal intensive care units, Dr. Felix Richter notes that few have leveraged this technology to conduct rigorous observational studies on patients.” “Our study demonstrates that deploying AI-powered algorithms on cameras monitoring infants in the NICU enables timely detection of neurologic changes, potentially facilitating prompt interventions and improved patient outcomes.”
The analysis team was astonished by Pose AI’s exceptional performance across diverse lighting conditions (day versus night). evening vs. In a study of infants undergoing phototherapy, researchers examined bilirubin levels from novel and varied perspectives. Surprisingly, they found that the Pose AI motion index was closely tied to both gestational age and postnatal age.
While this approach does not supersede the expertise of doctors and nurses in a neonatal intensive care unit (NICU), it remains crucial. This technology quite effectively augments existing data by providing a constant stream of information that can then be leveraged in a specific scientific setting, noted Dr. Richter. We anticipate a futuristic healthcare infrastructure where cameras continuously surveil newborns in neonatal intensive care units (NICUs), utilizing artificial intelligence to generate neuro-telemetry strips akin to those tracking heart rate and respiration, with real-time alerts for changes in sedation levels or cerebral dysfunction indicators. Clinicians may evaluate movies and AI-generated insights as needed, thereby offering a user-friendly and easily interpretable tool for bedside care.
Despite the acclaim surrounding the study, researchers also highlighted the limitations of the examination, including the fact that AI models were trained on data collected from a single institution, implying that the algorithm’s performance and neurological predictions must be evaluated using video data from multiple institutions and camera types. The analysis team intends to verify this knowledge in additional NICUs and design scientific studies that will evaluate the impact of this development on care practices. Researchers are also investigating the potential of this technology for various neurological conditions, with plans to extend its application to adult populations as well.
“At Mount Sinai, our commitment is to harnessing the potential of emerging synthetic intelligence technologies to drive innovative healthcare solutions for our patients,” said Girish N. Dr. Nadkarni, System Chief of Knowledge Integration and Digital Medicine, Director of the Mount Sinai Scientific Intelligence Center, and Co-Director of the Charles Bronfman Institute for Personalized Medicine. “Artificial intelligence instruments are revolutionizing healthcare across the Mount Sinai Health System, with tangible benefits including reduced patient stays, lower hospital readmission rates, enhanced cancer diagnosis and targeted therapy, and real-time care tailored to individual patients’ needs based on physiological data from wearables.” “We’re thrilled to introduce this cutting-edge, non-invasive, and highly secure AI technology to the NICU, poised to significantly improve outcomes for our most vulnerable and precious patients.”
