Recently, a woman with late-stage breast cancer was admitted to a city hospital with fluids already filling her lungs. She was attended by two doctors and got a radiology scan. Additionally, the hospital’s computer s were able to read her vital signs and projected a 9.3% probability that she would die in the course of her stay. When Google’s turn came, its new algorithm analyzed 175,639 data points and suggested that her death risk was 19.9 percent. The woman later died in several days.
Google AI published the disturbing account of the unidentified woman’s demise in research highlighting the healthcare power of neural networks. Neural networks are AI software that is good at utilizing data to automatically improve as it learns. Google has recently made headlines in the medical community for creating a tool that could be used in predicting numerous patient outcomes such as their odds of re-admission, how long they may stay in hospitals and the chances that they may soon die.
Google ’s capability to go through data that was previously out of reach impressed medical experts greatly. The neural net was fed with the unruly information and, in turn, produced or generated the predictions. What’s more, the technology was able to perform the task far much faster and accurately in comparison to existing methods.
Doctors, hospitals and other healthcare providers have been trying for several years to better utilize large volumes of electronic health records among other patient data. When more information is shared at the right time, it could aid in saving lives significantly. It could also assist medical workers in minimizing the amount of time they spend on paperwork and instead spend more time providing care to patients.
Nigam Shah, a co-author of Google’s research paper, said that utmost 80 percent of the time spent on current predictive models is channelled to the “scut work” of making the data more presentable. However, he pointed out that Google ‘s technique steers clear from doing so.
AI chief Jeff Dean said that Google’s next undertaking involves taking the predictive system into clinical settings. Dean’s health research department or unit also referred to as Medical Brain, is currently developing several AI tools that can forecast symptoms and disease with an accuracy level that is being met with both alarm and hope.
Software, particularly in the healthcare space, is widely coded through hand nowadays. On the contrary, Google ‘s method, where machines learn to break down data on their own, can catapult everything else. The machines can comprehend the problems that are worth solving.
Jeff Dean foresees the artificial intelligence (AI) system steering doctors to certain diagnoses and medications. Another researcher at Google said that the existing models are prone to missing apparent medical events such as whether a patient had undergone surgery previously. The unidentified researcher added that existing hand-codded models are a massive roadblock in the healthcare sector.
Despite the optimism surrounding Google’s capability, harnessing artificial intelligence to boost healthcare outcomes is still a major obstacle. Although other companies like IBM’s Watson unit have endeavored to apply AI into medicine, results have been mixed.
Source Bloomberg
