It is said that nothing is a matter of life and death except life and death. But people are hanging in-between - those who have fallen into a coma, lost all motor functions and no doctor could guarantee their recovery. Well, human doctors may not be able to ascertain what the future holds for these patients, but machine learning algorithms might be able to do it soon, going by what a group of Chinese neurologists are trying to achieve. Researchers at the Academy of Sciences and the PLA General Hospital in Beijing are working on a machine learning system that can predict whether a patient will ever wake up, reports Futurism. Developing such a tool involves feeding functional MRI data from thousands of coma patients to the AI system. Humans are doing the same thing but can never hope to be fast enough or take in so much of data. In contrast, machines do the scanning and matching at lightning speed. Also, looking at numerous parameters, a better prediction is made regarding a patient's chances.
It may seem a horrifyingly tricky territory for artificial intelligence to get into, but researchers argue that human knowledge is not doing much better. In fact, human fallibility has triggered the rapid development of self-driving vehicles where AI is deciding life or death for both passengers and pedestrians. The idea is that humans are too prone to causing accidents, influenced as they are by emotions and biases. AI is expected to fare better, being objective and able to access data sets without interference from feelings. This, of course, depends on whether the AI system in question has been built to be not too human.
But let us get back to the medtech innovation which seems to be working well. The algorithm is reportedly 88-90 per cent accurate, and the Beijing researchers in charge told the South China Morning Post they had used it to evaluate more than 300 patients. According to one of the doctors on the project, patients' families are given the AI score as part of the overall counselling and discussions but are also told such predictions should only affect 20-50 per cent of their decision and plans. It is because the system is not 100 per cent accurate. Moreover, the algorithm can only assess a person's brain functions and cannot account for other factors.
All these have raised many disturbing questions. How will the AI score impact human decision? Will it result in stopping the treatment and withdrawal of life support? Or will the medical care continue for an indefinite period? Some doctors not involved in the research say the error rate of humans are higher than what machines predict, but it is just as well that the last word is not left to technology alone.