A remarkable study in Lancet’s eClinicalMedicine has shown that artificial intelligence (AI) can identify serious neurologic changes in newborns in the neonatal intensive care unit (NICU) using only video data. Crafted by a team of skilled clinicians, scientists, and engineers at Mount Sinai, this cutting-edge method could transform how NICUs monitor neurological health.
The Challenge of Neurologic Monitoring in NICUs
Each year, over 300,000 newborns are admitted to NICUs across the United States. Many face risks of sudden neurologic issues. While heart and lung functions are continuously monitored, neurotelemetry has remained a challenge despite decades of research. Present methods depend on periodic physical exams that can sometimes overlook slight changes in a baby’s neurologic state.
The Role of AI and Computer Vision
The innovators at Mount Sinai used a deep learning technique known as “Pose AI” to follow infant movements through video data. This approach, popular in areas like sports and robotics, tracks anatomical landmarks using video feeds. Pose AI was trained with a vast dataset, covering over 16,938,000 seconds of footage from 115 infants at Mount Sinai Hospital. These babies were consistently monitored with video EEG from February 2021 to December 2022.
Accuracy and Versatility
Pioneering findings revealed that Pose AI could precisely track infant movements and predict two crucial neurological states: sedation and cerebral dysfunction. With impressive accuracy, the AI achieved scores between 0.87 to 0.91 for sedation and 0.76 to 0.91 for cerebral dysfunction. Various factors such as lighting or camera angles, be it during day, night, or under phototherapy, didn’t significantly impact these predictions.
Clinical Implications
This AI-driven technology offers a minimally invasive, efficient way to continually monitor neurologic health in NICUs. By delivering instant insights, it could expedite medical responses and potentially improve infants’ outcomes. While it’s not intended to replace traditional assessments by physicians or nurses, it enhances these evaluations by providing constant feedback that can guide healthcare actions. The vision for its future includes systems where cameras always watch over infants, creating a constant flow of neurologic data, much like heart or breathing monitors, with alerts for shifts in sedation or brain function.
Future Directions
Though initial results are promising, further validation is essential. Since models were trained using data from a single institution, it’s necessary to test these models elsewhere and with varied video equipment. Plans are underway for more testing in other NICUs and developing clinical trials to see how this tool impacts patient care. There’s also potential in adapting this technology for other neurological conditions and even exploring its use in adult patients.
In closing, blending AI with computer vision in NICU environments marks a significant leap forward in identifying and managing neurological changes early in newborns. This technology could notably improve patient outcomes, offering a continuous, precise, and comprehensible monitoring solution. It promises to address an existing gap in neonatal care, heralding a new era of medical monitoring and intervention.
Leave a Reply