Israel’s biggest hospital has hooked up military innovators, startups and defense companies to a cloud with patient data, and invited them to a Shark Tank-style competition, aimed at making a leap in coronavirus treatment.
The race is underway to build the best artificial intelligence tool to crunch the data of past coronavirus patients in order to save the lives of future patients.
A panel of 15 judges from Israel, America and Canada will hear pitches from finalists on Thursday and decide which Israeli team has won Sheba Medical Center’s Corona Data Challenge.
“It’s all about artificial intelligence, about predicting ahead of time what’s going to happen to patients, so doctors can care for them better,” Eyal Zimlichman, Sheba’s chief medical officer, told The Times of Israel.
Data scientists from the Israel Defense Forces, Israel Aerospace Industries, Elbit Systems, Facebook, and a dozen other companies and organizations are waiting to find out if they have made the shortlist of five teams that will pitch during the online final.
“We realized at the start of the crisis that data is going to play an important role, and as many organizations, including the army and defense industries, started asking for data,” said Zimlichman. “In one of the only initiatives of this type in the world, we built a system to upload every patient’s data, anonymized, to the cloud, every 24 hours.”
It contains every piece of medical information recorded on more than 500 patients from their entire hospitalization — down to every blood test and every scan. Some 30 groups of data scientists have accessed the data.
Now, the team that judges deem to have produced the smartest artificial intelligence tool will be selected to develop and implement it at Sheba, at the hospital’s expense, and also test it at New York’s Mount Sinai Hospital, one of the biggest teaching hospitals in America.
Competitors have strong data science credentials, but lack medical expertise, so Sheba doctors have been delivering weekly webinars to give medical context to the data.
Teams have focused on how to predict which treatment best fits different patients. Some have addressed questions like the best method of ventilation, while most have worked on AI tools to explore the topic that continues to perplex doctors, namely, how to predict which patients are headed for a sudden deterioration.
Zimlichman said: “The most common question is to try to predict which patients are going to deteriorate in to a severe COVID-19 state, because we still don’t have a clue why some patients become severe while others don’t.
“If there were algorithms to predict risk level, we could know where patients should be located and what treatment should be prepared at different points. It’s about predicting outcomes and getting early warning of deterioration.”