Reducing neonate mortality rates with AI and Edge computing
It’s 9 PM on a Wednesday, and she’s in the Emergency Room. Monitors are attached to her arms, legs, chest and midsection, and an IV port is in her arm. A flurry of calm but directed individuals appears from between the curtains. They leave the room with pleasant goodbyes, but their eyes and glances with their peers tell a different story. Then she hears the words: “HELLP Syndrome,” “pre-eclampsia,” “about 4 lbs.,” “it’s not that small,” “we’ll take good care of you–both.” The gurney is rolling toward an operating room. Say the alphabet backward, starting with z. She delivered her 3 lb. 7 oz. baby eight weeks early. Forty-eight hours post-birth, her child was jaundiced—a complication of the bloodstream infection sepsis. Thus, begins the adventure into the unknown.
–Reflections from a preemie’s parent
Premature children, especially those under 1500 grams (3.3 lbs.), are susceptible to significant health challenges. Their immature neurological, cardiac, pulmonary, renal and immune systems can make them vulnerable to life-threatening complications. And when issues surface, physicians and healthcare staff must respond quickly. Sepsis is one of the significant health challenges. Without rapid intervention, up to fifteen percent of premature infants perish due to sepsis complications. Timely sepsis diagnosis and treatment can be the difference between heartbreaking or successful outcomes.
Patient monitoring in a NICU environment generates exceptional amounts of structured and unstructured data. It is collected in varying time intervals formatted and typically retained in multiple, heterogeneous systems and applications. Despite volumes of data and high-touch healthcare provider-to-patient ratios, diagnosis of sepsis can take several hours after its onset. When it comes to NICU infants, every hour counts. Could AI help expedite the diagnoses and treatments for the most vulnerable?
For more than a decade, neonatologist Dr. David Van Laere has cared for some of the smallest infants in Australian and Belgian neonatal intensive care units (NICUs). The lessons he learned while treating infants afflicted with sepsis affected his life’s path. In addition to his practice as a neonatal intensive care specialist, he founded the Innocens Project. The organization was established at the University Hospital at Antwerp in Belgium with the mission to speed time to diagnoses and treatment of sepsis. Physicians and technologists combined their years of in-depth medical expertise and research with three years of patient data and sought to battle their opponents.
Dr. Van Laere was convinced the solution to swift diagnosis and treatment was buried within patient data—metrics, measurements and imagery captured during an infant’s NICU stay. At the bedside, I started monitoring at data trends. My research focused on signal analysis and trying to extract additional information from vital signs that related to complications, such as sepsis,” said Dr. Van Laere. He searched for a data and technology solution that securely managed and analyzed personal patient data, automated and modernized the diagnosis process, and predicted when an infant was trending toward sepsis and triggered an evaluation protocol. He discussed the challenge while biking weekly with his neighbor Dirk Claessen, a managing director with IBM Global Markets
Although Claessen worked with large clients in the oil and gas industry, his in-depth AI and predictive analytics expertise and IBM ecosystem knowledge allowed him to seamlessly find a solution to the healthcare challenge at hand. Thus, began the collaboration between Dr. Van Laere and Innocens, IBM, IBM Design Center in Munich and IBM Business Partners KPI Digital and Comsense Technologies. The collaboration resulted in an Edge computing solution powered by IBM Cloud that captures real-time data generated by the Internet of Things sensors. Data analysis is facilitated through the open-source Red Hat® OpenShift® Container Platform and IBM Watson® Studio on IBM Cloud Pak® for Data. “I felt that we needed to have both a research focus but a deployment focus at the same time. And then, we needed a trusted tech partner that could deliver on all the aspects of trustworthy AI. And so that’s the partner that I found with IBM,” said Dr. Van Laere. The IBM Business Partners and Innocens earned the distinction of the 2021 IBM Beacon Award for its integrated Data and AI solution.
The continuous cycle of current and historical patient data integrated with artificial intelligence, machine learning and data modeling capabilities provided the ability to identify infants at risk of sepsis up to eight hours before the standard protocol for monitoring and testing. Furthermore, the AI solution is 75% accurate in detecting severe sepsis with less than one false alarm per week. This rapid diagnosis combined with expedited treatment capabilities results in decreased mortality rates, quicker recovery and shorter hospital stays.
Dr. Van Laere and the Innocens team have just begun their AI and Edge Computing journey. The team plans to deploy the sepsis diagnosis and treatment solution within other NICU hospitals and systems. “Basically, the sky’s the limit. We can try to build a model to predict brain injury in preterm infants. We could obtain additional insights on chronic lung disease or retinopathy eye disease that is due to prematurity,” Dr. Van Laere concludes. “I feel that exciting times are ahead for us in neonatal intensive care unit in Antwerp and at Innocens.”
IBM Health Forum features over 25 client and expert technology sessions like the Innocens Project. Here the experiences of healthcare organization as the integrate AI solutions to improve operational efficiencies, strategies for growth and patient experiences. Register for the replay here.
To learn more about the Innocens Project, view Dr. Van Laere’s IBM Health Forum session replay “Every hour counts: Catching Sepsis early in NICU Infants.”
Visit Journey to AI for additional resources
The post Reducing neonate mortality rates with AI and Edge computing appeared first on Journey to AI Blog.