Technology giant Alphabet Inc.’s Google (GOOGL) has been attempting to make an entry into the health care space for a while now, and the efforts seems to be paying off. Google claims to have created a system that is capable of forecasting a variety of outcomes for patients, including the duration for which people may need to be hospitalized, their chances of readmission and their chances of death. Called the Medical Brain, this breakthrough could give Google a completely new market to explore. (See also: How to Play the Coming AI Revolution.)
Bloomberg reports a case study of a woman having late-stage breast cancer who was given a survival chance of 9.3% by the hospital’s standard computing methods, while Google’s predictive analysis gave her a 19.9% chance of dying during her hospital stay. The patient passed away within a few days, firming up Google’s claims for offering a better prediction mechanism by its system.
In the May edition of the scientific journal Nature, Google's team described its predictive methodology: “These models outperformed traditional, clinically-used predictive models in all cases. We believe that this approach can be used to create accurate and scalable predictions for a variety of clinical scenarios.” The research highlights the use of neural networks in the field of health care. A neural network is a form of artificial intelligence (AI) software modeled on the human brain and nervous system that relies on using data to automatically learn and improve at identifying underlying relationships.
How Google’s Tool Works
Medical practitioners, hospitals and other health care providers have been struggling for years to better maintain and summarize medical data for a patient. However, despite the use of advanced data storage systems dedicated to hospital use, success has been varying.
Available reports indicate that Google’s system for such predictive analysis works on sifting through tons of data points to arrive at the inference. In the above case, Google’s algorithm analyzed 175,639 data points to make its conclusion. Google’s capacity for reading data in a variety of forms—including handwritten notes saved as PDFs, old charts and medical reports—combined with its processing speed is the real game changer. The algorithm also demonstrates which data points were most useful in reaching the conclusion.
While present-day predictive models spend around 80% of its time on data scouting and presentation, Google’s approach avoids this bottleneck. (See also: Google's DeepMind Gets Access to NHS Patient Data; Controversy Ensues.)