Every company is using AI to make tools which are useful for the people. Google has been working a lot in this field trying to create new tools which will benefit the people. Recently, the Medical Team of Google created a new AI which will predict death risk of hospital patients. The tool is said to be working better than the hospitals which will predict if the patient will die in the next 24 hours of being admitted.
Moreover, the tool is said to be 95% accurate which is far better than any existing systems. The system will use data about the patients, their age, gender, signs from previous reports, lab results, and diagnostics report to predict the risk percentage for the patients. The data can be fed through the doctor notes or PDFs, unlike other systems. It will identify all the information needed from them and predict.
In a paper published by Google in May, it said – “The prediction model outperforms all the other traditional models present. The new approach can be used to create scalable and accurate predictions for various clinic scenarios.
Artificial Intelligence has become smarter and more powerful these days. They make themselves better using the Machine Learning.
Google’s New AI
The new AI system is developed by a team of scientists from Stanford, UC San Francisco, and the University of Chicago. Google took control of the AI system and fed it with 216,221 adult patient details obtained from two US medical centers. The AI tool got 46 billion data overall which is used to improve its own prediction system. While learning, the AI tool became self-sufficient with many terms of life and death. It was then able to predict how likely or unlikely the patient is towards death.
In a testing phase, the tool was tested on a patient with metastatic breast cancer. Google’s tool predicted a 19.9% chances of her death in the hospital whereas the Hospital tool predicted 9.9%. The patient passed away in the hospital within 2 weeks which proved the efficiency and correctness of Google’s AI. The AI used over 175,639 data points of the patients from various medical records. The explanation given by Google on the correctness of the tool is that “The previously available tools used some of the data available EHR rather than the whole data. But, our new tool uses all the data including handwritten notes present in the EHR”.
The AI system is not the first step Google has taken in the healthcare field. Earlier, Deepmind partnered with Department of Veterans to feed 700k records for predicting changes in the condition of patients. They are also working with a Voice recognition tool for the clinic but it depends on the accuracy. In the medical field, even a minor mistake can prove fatal. But, the tool is designed to analyze any type of notes and thus removes the human error.
The AI tool is not limited to predicting the death or live status of a patient. It can also predict the duration for which the patient will be in the hospital and also chances of being re-admitted. Interestingly, the tool is can accept nearly every type of data fed into it. This enables the tool to get more data to increase the accuracy of the system.