Artificial intelligence can now represent the difference between life and death for patients in a coma. In Beijing, China, human doctors deemed that at least seven patients in a vegetative state had no chances of regaining consciousness. However, an artificial intelligence algorithm developed by the Chinese Academy of Sciences and PLA General Hospital, predicted that these patients would awaken within a year. An automated instrument predicting human life seems impossible, but that is exactly what the program did. All seven patients regained consciousness within that year, just as the AI program predicted.
This AI system, developed after the past eight years of machine learning and research, touts an 88 percent accuracy on prognostic assessments in patients with comas. The algorithm works by analyzing functional magnetic resonance imaging (fMRI) scans of patients’ brains. Dr. Song Min, associate researcher and first author of the research study published in August 2018, notes that their “machine can ‘see’ things invisible to human eyes” since they can consider pictures at a per-pixel level. In this case, the artificial intelligence program can scrutinize minute neural activities that conventional evaluation methods cannot. Indeed, the system can look very closely at different regions of the brain controlling hearing, vision, motion control, and verbal capability.
Doctors conventionally test patients in vegetative states by their responses to sensory stimulation, which may lead to a misjudgment of a patient’s condition. The problem? Patients in comas may have an actively working mind, but are blocked from communicating with the world around them.
So what is the future for this type of technology? To find out, I went to the source. Dr. Bharat Biswal, the very inventor of the resting state functional MRI, the technology now being used for the program in Beijing, is a distinguished professor at NJIT in the biomedical engineering department. While most researchers use MRIs to study the brain performing tasks, Biswal found that even at rest, the brain holds valuable information that doctors can use to diagnose various brain diseases.
Dr. Biswal has high hopes for the future. He notes that the resting state fMRIs can help determine which brain “functions have been affected or which ones have stayed intact” in patients suffering “strokes, traumatic brain injury, or in infants that do not have the mental capability to perform tasks well.” Biswal also states that “machines can process a lot of information at once,” so machine learning is only going to improve and help more people than ever.
This isn’t the only case in which AI may supersede the abilities of medical doctors. Stanford University developed an AI program that could categorize a skin condition with just a picture, and often with more sensitivity and accuracy than a doctor could.
However, this isn’t to say artificial intelligence systems are going to replace doctors. Indeed, more often than not, the AI assessment in the Beijing project matched that of doctors’. Min notes that their program “will never replace doctors, but is simply a tool to help doctors and families make better decisions.” The program hasn’t been correct every time either. In one case a 36-year old man scored very low on both the doctor’s and the algorithm’s scale, predicting he would never wake up, but he regained consciousness within a year. But the data set is only growing and the machinery only improving.
There are certainly varying opinions as to the extent to which AI can and should be used. AI programs may never have the same level of sensibility as regular doctors in accounting for outside factors in diagnoses or treatment of patients. No program can take everything into account. For this, Dr. Biswal believes that “AI machinery can complement physicians,” providing more input and strengthening the diagnoses or treatments. On the other hand, computer science NJIT associate professor Dr. Zhi Wei states that we are moving towards a better future with the development of fully-implemented AI in medicine, as machines can “do even better than humans at a lower cost.” This may even point to cheaper, more accessible health care for the future.
In any case, as Dr. Biswal states, “new knowledge is always useful. It’s only what we do with it that may become an issue.”
‘
