How Artificial Intelligence Could Diagnose Schizophrenia
Of the 3.2 million people in the US who are diagnosed with Schizophrenia, nearly all of them share one comment experience in the treatment process: a psychologist or psychiatrist consults the patient, analyzes potential mental illness markers, and delivers a diagnosis with the best treatment available for that patient.
As artificial intelligence (AI) continues to advance, it does so not only in self-driving cars, or your next virtual assistant, but also in diagnosis and treatment of schizophrenia and other psychotic disorders.
The technology is exciting and remarkably innovative, utilizing improvements of a simple technique used by psychologists and psychiatrists across the world: talk therapy.
Since Sigmund Freud first sat a patient on his classic psychoanalytic couch, patients and doctors have relied primarily on therapy sessions to develop an idea of what may be happening within the patient’s mind, and allow the doctor to recommend a basis of treatment with which to proceed. The problem with this dated system arises when a doctor makes an ill-informed decision based on only their understanding of a patient, which is often incomplete.
A hallmark of psychosis, one of the precursors to schizophrenia, is the presence of disjointed thoughts and speech. It is these abnormal patterns that a psychologist can identify in order to narrow a probable diagnosis and begin treatment of a disorder. Of course, not all diagnoses are accurate, and anything a human can do, a machine can do better. It is this juncture where AI can bridge the gap between a psychologist’s recommendation and a comprehensive diagnosis.
An AI program from researchers at Columbia University, the New York State Psychiatric Institute, and the IBM T. J. Watson Research Center differentiated — with 100% accuracy — between patients susceptible to psychosis who would go on to develop psychosis and those who would not; traditional psychologists, on the other hand, averaged an accuracy rate of only 79%. Similar to the psychologists, the program intensely analyzes speech patterns to discern patients developing psychosis from those who were not. Guillermo Cecchi, a biometaphorical-computing researcher for IBM notes that, “as an interviewer, if [the psychologist’s] mind wandered briefly, [they] might miss [a sign of developing psychosis]. But a computer would pick it up.” The powerful analysis of an AI system cannot be matched by a psychologist, the system eliminates human error to provide an impeccable diagnosis.
Not only may AI make advances in the world of talk therapy, it could also run analyses of Functional Magnetic Resonance Imaging (fMRI) scans to examine a patient’s brain for areas of activity related to schizophrenia.
Psychosis causes a patient to feel distant from the world around them, leading to a sharp decline in emotional reactions, similar to the lack of reactions experienced by those affected by an Autism Spectrum Disorder (ASD).
Dr. Mitsuo Kawato of the ATR Computational Neuroscience Laboratories in Kyoto, Japan has been using an artificially intelligent system with an fMRI to scan brains of ASD patients and compare them to healthy brains to establish the differences in brain connections. Not only is this strategy remarkably effective in ASD cases, the same “thought markers” often apply to psychosis, and thus schizophrenia.
By pinpointing the lack of emotional response in the brain, an artificially intelligent system can alert a doctor to psychotic symptoms, and calculate the probability of a patient developing schizophrenia by analyzing the fMRI and speech pattern data that the program has gathered. The sum of the information a computer gathers allows the system to develop a full mental profile of a patient, which can then be shared with a psychologist or psychiatrist for double-checking and administration of care.
Patterns of disjointed speech may appear in very mild form years before a patient develops schizophrenia, and the breaks may be so subtle that even a trained psychologist could miss them. It is at this point that an artificially intelligent system could be utilized to save the patient and his/her family time, money, and energy in searching for care by identifying a patient’s problems early. The average onset age of schizophrenia (in genetic cases), is 18–25 years old. If a possibly psychotic patient participates in one therapy session and undergoes fMRI testing, the system would be able to diagnose the patient based on those two analyses and allow the doctor to begin to formulate the treatment strategy most effective for that individual.
Dr. Kawato’s program’s ability to distinguish future psychotic patients from those who will not develop psychosis (and thus likely schizophrenia) is a revolutionary advantage to patients and doctors in the field of psychology: doctors will have a new tool of diagnosis and treatment recommendation, and patients will have access to a world of help from their local doctor.
A wealth of knowledge at their fingertips is what every doctor desires when treating a patient, and a system of programs could provide just that. As opposed to doctors manually sorting through long lists of files regarding different tests that have been run on a patient (EEG or blood work), an intelligent computer system could actively take into account all of the results of the different tests to give a preliminary diagnosis to the patient’s doctor, with which the doctor could determine a course of treatment.
This comparative system could apply to an individual patient, and many different computers could link their experiences to help patients around the world based on very recent cases. If a patient came to a psychologist complaining of what the doctor thought to be a unique set of circumstances, the doctor could consult his computer’s database to recognize what may have worked for similar patients in the past, as well as any new treatments that could be effective in the case. Not only will the system help patients every day, it will drive innovation in the mental health field to create solutions to any possible situation in every possible patient.
Of course, a computer listening to every word and every thought a person has, and being able to share that information with the world, is a double-edged sword. Patient privacy plays a major role in all health care, especially mental health care, and some patients may be turned away by the prospect of a sentient computer having their mind on file. Even if patients were to accept treatment aided by an artificially intelligent system, no system is currently able to provide a comprehensive diagnosis by itself; a doctor would be necessary to analyze the system’s results and administer effective treatment to a patient.
The world of mental health care has been fairly steady; few advances in diagnosis or treatment have come forward in recent years, and the benefits of a supercomputer doctor are evident across many fields of medicine. With artificially intelligent technology, future generations will enjoy earlier detection of their illness, collaborative treatment, and prevention of symptoms if only a system of such brilliance can be implemented ethically and effectively.
