In a breakthrough development to serve COVID-19 patients, Addenbrooke’s Hospital at Cambridge and 20 other hospitals world over along with tech giant NVIDIA have used AI to predict the oxygen needs.
In fact, the pandemic sparked research and led to the development of AI tool to anticipate the need for extra oxygen that a patient may require in the first days of hospital stay using data available from four continents.
Known as Federated Learning, the technique used an algorithms to examine chest X-rays and electronic health data of patients admitted in hospitals with Covid symptoms.
For strict confidentiality of patient data, the data was modified to be fully anonymous and an algorithm sent to each hospital for the data to remain intact at its original source. On learning of data from the algorithm, the analysis was compiled to develop an AI tool to anticipate the oxygen needs of COVID patients that are hospitalized anywhere in the world.
The finding published in Nature Medicine is one of the most diverse and largest clinically federated study till date.
Meanwhile, to establish the accuracy of the AI tool, it was investigated in several hospitals across five continents. The results demonstrated the prediction of oxygen needed within 24 hours of an arrival of a patient in the emergency department had 95 percent sensitivity and specificity of more than 85%.
Importantly, the technique has the transformative power to bring innovations in clinical workflows. The continued work with the technique demonstrates that this type of global collaborations can be repeated and can be improved for efficiency so that the need of clinicians can be made to handle challenges and epidemics in the future.