Diagnosing a psychosis with big data: IBM’s challenge

Psychoses, including schizophrenia as a representative example, tend to develop in adolescents’ mind. Early detection and proper intervention are crucial for the better prognosis of psychotic patients. However, an exact diagnosis is often difficult.

To address the variance of diagnostic criteria, Robert Spitzer and his colleagues in American Psychiatric Association established an operationalizing diagnostic criteria, the DSM. Although DSM had been traditionally used also before Spitzer, it was scrapped and entirely revised. Now, DSM-5, the newest version of DSM, gives us relatively reliable and valid standard of psychiatric diagnosis.

So, has the diagnostic methodology evolved in recent years? There are still several controversy on this matter, unfortunately.

Recently, IBM entered the labyrinth to seek the goal.

It is well known that some mental symptoms appear in the prodromal state of psychosis. They include thought disturbance, cognitive deficit, hypersensitiveness, and incoherence in the flow of thinking. The researchers in IBM is seeking some particular patterns suggesting such symptoms in their natural conversation. Specifically, patients are asked to reply some open-ended questions. Their answers are recorded. The content of their speech is assessed based on some special algorithms. Finally, the program detects unnatural thought or flow of thinking in the records.

IBM: Word Analysis Proves Effective Pre-Psychosis Diagnosis

This attempt is fascinating. We, psychiatrists, do an interview with patients in the daily clinical work. Through the conversation, we sense the oddness or abnormalities in the patients, aiding us to diagnose them. Some symptoms are easy to detect, but others are hardly identified. Sometimes we feel difficult to explain why we diagnosed the patients as so.

Examination with a computer includes no vagueness. It will contribute to the precise diagnosis. On the other hand, many patients cannot be identified that they need psychiatric help. In such cases, clinical practitioners tend to get in touch with the patients so that they can intervene the patients when it is required. Computer diagnosis cannot deal with patients who are in the grey zone.

For the reason above, I do not think that the computer diagnosis will replace psychiatrists until the near future. In contrast, diagnostic criteria for some mental disorders will be sophisticated using this digital diagnostic system. Current criteria are described by Spitzer and relevant clinicians. But with big data, more specific symptoms mentioned in natural conversations will be identified. I think it is possible.

Therefore, I am not afraid that the computer will get rid of psychiatrists’ chairs soon, but the DSM-6 will be written by the computer.