Google can accurately predict when a patient will die using AI

byrawpixel

Google is entering the medical field using artificial intelligence (AI), and let AI predict the probability of death by cardiac disease by machine learning of patient's dataResearch has also been conducted. Such medical technology is being improved as of June 2018, and the Los Angeles Times summarizes how Google's algorithm reduces waste of modern medicine and improves efficiency.

Google is training machines to predict when a patient will die
http://www.latimes.com/business/technology/la-fi-tn-google-artificial-intelligence-healthcare-20180618-story.html

The algorithm to estimate the patient's death probability has been developed before Google entered the medical field, but its accuracy was low. For example, a woman from a late stage breast cancer went to a hospital and received a radiological examination as recommended by a doctor, and a hospital computer analyzed the examination result, the probability of death of a female patient was estimated to be 9.3%. However, after that, when Google's algorithm examined 175,639 kinds of data points on women, the risk of death of female patients was estimated to be 19.9%, which is more than twice as high as the first time. A few days later, the female patient died.

A Google algorithm that predicts the probability of death of a patient more highly than conventional algorithms is a type of AI software announced in May 2018. It is particularly advantageous in that data can be automatically learned and performance can be improved, and accordingly, Google notifies "how long the patient will stay in the hospital?" And "the possibility of re-hospitalization of the patient" It will be able to predict.

Google develops technology to predict "what next will happen" to the patient who visited the hospital - GIGAZINE


The reason why Google's AI software was impressive for medical staff seems to be that the neural network can capture all data that was not used before (such as notes embedded in PDF) and can execute prediction . And Google's algorithm is much faster and more accurate predictable of the patient's future than the conventional technology.

byrawpixel

Hospitals, doctors, and health care providers have striven to stock electronic health records and other patient data. By sharing more appropriate information at the right time, it will be possible to provide better medical care. However, the previous methods, such as mining medical data, were troublesome as it took time and costly to produce results.

Nigam Shah of Stanford University asserts that "80% of the time spent in traditional patient prediction models is spent only to make the data easier to view". Google's approach seems to avoid this "wasted time".

Also, while most of the software used in medical care is coded manually, Google takes the approach of letting computers learn data analysis methods automatically. Vic Badger, former president of Google's parent company Alphabet's healthcare department and managing director of investment company Foresight Capital, said, "What's the obstacle for Google's approach to solving problems? "I am doing enough experiments in order to know exactly the direction of fruitfulness" and I emphasize that there is no problem with Google's method.

byJESHOOTS.COM

Of course, there are problems that stand out before the fusion of AI and medical care by Google. "Google and other technology leaders will have dominant power by exclusively utilizing all the data we generate," said Andrew Bart of data company Immuta Mr. Mr. Bart and pediatric oncologist Samuel Volchenboum are concerned that Google will dominate medical data like Google dominates online advertisement and the government firmly holds the government to avoid it We insist that we should. Nonetheless Volchenboum is convinced that the algorithm will help save a lot of life and money, he says that health record hopes to mix with other statistical data and provide new added value.

Google uses AI to create various medical-related tools, which can be licensed to clinics or sold as a type of service through the cloud computing department. However, in order for Google to commercialize the AI ​​tool, we need to acquire a wider range of data. Google is able to purchase such data, but it is obvious that regulators and consumers will not think so well, and it will be a future focus on how to clear this problem I will.

Google believes that it is premature to include medical-related AI tools in business models, and even in Google I / O held in May 2018, Lily Peng who presented a tool to detect cardiac risks "We want to emphasize that this tool is really in its early stages," he said.