Google's Medical Brain team uses AI to predict patient's health and even death risk

According to Google, its AI is much better and faster at predicting death than hospitals

Google’s 'Medical Brain' team has now integrated artificial intelligence to predict the death risk among hospital patients. The results have surprised medical experts by showing that it has quite a high degree of accuracy than a hospital’s in-house warning system. It certainly looks like the advances by the Medical Brain team could help the company in the healthcare business.

According to a Bloomberg report, Google’s health-care potential of neural networks is a form of artificial intelligence software that’s particularly good at using data to automatically learn and improve.

A nurse treats a patient wearing the 3D therapeutic virtual reality headset. Image: Reuters

A nurse treats a patient wearing the 3D therapeutic virtual reality headset. Image: Reuters

The tool that can forecast a host of patient outcomes, including how long people will stay in hospitals, their odds of re-admission and the death risk probabilities.

Google's team on 8 May this year, published its findings in a paper on Nature  titled 'Scalable and accurate deep learning with electronic health records.'

In a case study of a woman with metastatic breast cancer, Google utilised its algorithm and 24 hours after she was admitted, Google predicted a 19.9 percent likelihood of her dying within the hospital. This is was in complete contrast with the 9.3 percent estimate that the hospital’s augmented Early Warning Rating (aEWR) predicted. The patient succumbed to death within 10 days of admission.

The incident was rather harrowing but has convinced and impressed medical experts. Mostly because Google's systems have the ability to go through the data which was previously out of reach, for instance, notes buried in PDFs or scribbled on old charts. Manual reach takes tremendous amounts of time, and the hospital's internal systems cannot sift through and comprehend this information.

Google AI helping in detecting defect in the cornea. Image: Google

Google AI helping in detecting defect in the cornea. Image: Google

What is interesting is that not only does the neural network collect all this information for predictions, it also links back the predictions to which records it drew its conclusions from. This makes it look more credible. It has proven to be way faster and more accurate than existing techniques.

In the healthcare sector, most of the information is entered manually even now. Google’s approach, where machines learn to parse data on their own "can just leapfrog everything else," Vik Bajaj who worked at Alphabet's health-care arm, told Bloomberg.

Using AI in healthcare does, however, remain a challenge because it might create fears of machine learning predicting and having an incredible amount of say over who will get what care, and will die when.

Other companies have tried to apply artificial intelligence to medicine, like IBM's Watson Unit, but have had a tough time integrating it.

According to the report, it is not sure if the tech giant will use the neural networks as a business model, however, at the Google I/O this is year, in May, Medical Brain's Lily Peng emphasised how it outmatched humans at spotting diseases.




Top Stories


also see

science