Could Google tell you when you'll die?

Google may one day be able to predict when you'll die years in advance.

The firm has created an AI that it claims is 95 per cent accurate in predicting whether hospital patients will pass away 24 hours after admission.

This is around 10 per cent better than traditional models.

To make its predictions, the software uses data such as patient's ethnicity, age, gender, previous diagnoses, lab results and vital signs.

But what makes it so powerful is that it includes data previously thought out of reach of machines, such as doctor notes buried in PDFs or scribbled on old charts.

As well as death, AI can also predict unplanned re-admissions within 30 days and  probable length of stay at a hospital.

The system is still in its infancy, but Google believes it could someday be used to predict death far longer in advance.

Google has created an AI that it claims is 95 per cent accurate in predicting whether hospital patients will pass away 24 hours after admission. These graphs show a Google AI's performance in predicting mortality (solid line) compared to baseline computer models (dotted line) at two hospitals            +4

Google has created an AI that it claims is 95 per cent accurate in predicting whether hospital patients will pass away 24 hours after admission. These graphs show a Google AI's performance in predicting mortality (solid line) compared to baseline computer models (dotted line) at two hospitals

The AI was developed in collaboration with colleagues at UC San Francisco, Stanford Medicine and The University of Chicago Medicine.

To test the system, Google obtained de-identified data of 216,221 adults, with more than 46 billion data points between them.

After studying the data, the AI was able to identify which words were associated closest with outcomes.

While the results have not been validated, Google claims huge improvements over traditional models.

The biggest benefit, researchers claim, is the ability for the system to use all types of data.

Nigam Shah, an associate professor at Stanford University told Bloomberg that as much as 80 per cent of the time spent on predictive models goes to making the data presentable.

'[With Google] you can throw in the kitchen sink and not have to worry about it,' Shah said.

In a written statement, Google research scientist and medical doctor Alvin Rajkomar MD said: 'When patients get admitted to a hospital, they have many questions about what will happen next.

'When will I be able to go home? Will I get better? Will I have to come back to the hospital?

'Predicting what will happen next is a natural application of machine learning.'

The software scans medical records, analysing hundreds of thousands of data points. In the case of a woman with breast cancer (records pictured), the hospital gave a 9.3% prediction she would die during her stay, compared to Google's 19.9%, which sadly proved more accurate

Google's tool can predict a range of patient outcomes, including the length of a patient's stay in hospital as well as their chances of readmission.

For each prediction, a deep learning model reads all the data-points in electronic health records, from earliest to most recent, and then learns which data helps to predict the outcome.

The final results showed impressive accuracy.

Where 1.00 is perfect, and 0.50 is no better than random chance, Google's AI scored the following: 0.86 in predicting if patients would stay long in the hospital, compared to 0.76 using traditional methods; 0.95 in predicting inpatient mortality, compared to traditional methods at 0.86; they 0.77 in predicting unexpected readmissions after patients are discharged, compared to traditional methods at 0.70.

The full findings of the study were published in the Nature partner journal Digital Medicine.

AI systems rely on artificial neural networks (ANNs), which try to simulate the way the brain works in order to learn.

ANNs can be trained to recognise patterns in information - including speech, text data, or visual images - and are the basis for a large number of the developments in AI over recent years.

Conventional AI uses input to 'teach' an algorithm about a particular subject by feeding it massive amounts of information.  

Practical applications include Google's language translation services, Facebook's facial recognition software and Snapchat's image altering live filters.

The process of inputting this data can be extremely time consuming, and is limited to one type of knowledge.

A new breed of ANNs called Adversarial Neural Networks pits the wits of two AI bots against each other, which allows them to learn from each other.

This approach is designed to speed up the process of learning, as well as refining the output created by AI systems.

Follow NEWS.am Medicine on Facebook and Twitter