How A New AI Diabetes System Outperformed Top Docs
spire Ventures pitted its adaptive artificial intelligence platform, A2I, against some of Europe’s best diabetes specialists in an observational trial and the A2I won.
A diabetes management system built on a new artificial intelligence platform proved it could offer people with diabetes a less invasive and more personalized solution for managing their blood glucose levels.
Diabeter, a Netherlands-based diabetes clinic and research center that Medtronic bought in 2015, is known for routinely achieving some of the best results in the world in terms of managing diabetes, but a new artificial intelligence-based system could potentially make the clinic’s diabetes management results even better. That was the conclusion of an observational study authored by Diabeter doctors who tested the Rhythm system developed by Tempo Health LLC, an Aspire Ventures portfolio company.
The closed-loop Rhythm system is built on Aspire’s adaptive artificial intelligence platform, A2I. Diabeter presented the findings this week at the annual Advanced Technologies & Treatments for Diabetes conference in Paris, France. The clinic said it frequently participates in technological studies that could potentially improve the lives of people with diabetes.
Through the A2I platform, Tempo developed Rhythm to forecast and manage blood glucose levels of people with diabetes, based only on non-invasive biometric sensors (stickers) and artificial intelligence. By leveraging personalized blood glucose prediction models that adapted to each of the eight patients who participated in the observational study, researchers found that in seven of the eight patients, the Rhythm system alone would have been able to achieve a 20% increase in time in range, and a 9% reduction in lows, as compared to the actual results achieved by active control tower monitoring by focused and experienced doctors and their diabetes teams using patient-activated remote monitoring.
A2I uses a vast number of algorithms for self-optimization. The technology is designed to use any type of data, from text to video to biometrics, and draw from a library of analytical components to assemble, optimize, and combine components from multiple algorithms to build the best possible adaptive algorithm.
Mike Monteiro, chief data science and innovation officer, told Qmed the Rhythm system will be classified as an artificial pancreas technology, but he explained how the system is different from other artificial pancreas technologies. Other systems in the artificial pancreas category are designed to act sort of like thermometers for the body, Monteiro said, whereas “we’re building a nest.”
“Not everyone is using the same predictive algorithm,” Monteiro said. “We’re automatically selecting the best individualized algorithm customized to each patient’s biometrics.”
So the Rhythm system, using the A2I platform, is designed to listen to the patient’s biorhythms and learn their behavioral patterns in order to figure out what their blood sugar needs to be and what the patient needs to do in order to keep their blood sugar level in that ideal range.
Dick Mul, a pediatric endocrinologist for Diabeter, said the trial results were encouraging, and offered a positive outlook for potentially improving diabetes treatment for people worldwide. “The good news from the trial results is that both in range increased and time in hypoglycemia decreased,” Mul said. “This can be achieved without the direct need for additional or constantly invasive continuous glucose monitoring devices, and might help to reduce the need for active manual remote monitoring by a clinician. This is important, as currently not all our patients will yet be able to use sophisticated technological sensor-augmented insulin pump systems.”
Monteiro said he was actually surprised at how well-received the system was among the Diabeter doctors. He anticipated more resistance from the specialists, and a mentality of “how can this do my job better than I can?”
“Pretty much all doctors, at the end of the day, really just care about outcomes for their patients,” Monteiro said. He added that the doctors who participated in the study also expressed a sense of relief at not having to do all of the data processing themselves in order to help their patients achieve better outcomes.
“Doctors are doctors,” he said. “They’re not data processors.”
By using technology like the Rhythm system, doctors can potentially spend more time focusing on patient care and less time analyzing data from a control tower.
Monteiro said the next step is to carry out a larger clinical trial an pursue FDA approval. The plan, he said, is to strip the functionality of the system into a couple different pieces in order to get through the regulatory process faster, but the company will also pursue approval in parallel of the full system.