A new neural network developed by Google based on deep learning algorithms is able to determine the approximate timing of death, based on the patient's medical parameters. Bloomberg reports.
The algorithm has been tested on several hundred thousand patients, including a woman suffering from end-stage breast cancer, whose lungs have already begun to fill with fluid. Computers analyzed the X-ray scan data and determined that there was a 9.3 percent chance of the patient dying during her hospital stay. However, the neural network, which took into account 175,639 parameters, increased this indicator to 19.9 percent. The woman passed away ten days later.
The neural network is able not only to predict death, but also other outcomes of diseases, as well as the length of stay in the hospital and the likelihood of readmission. The algorithm can take into account records in PDF files and old medical records. It also saves physicians up to 80 percent of the time they would normally spend on routine data preparation.
According to company representatives, Google's next step will be to implement its predictive system in clinics.