Diagnosing Schizophrenia with AI | Mind You

By Dr. David Laing Dawson

Marvin had me listen to a podcast about applying AI to the diagnosis of schizophrenia.

And it is true that AI, in conjunction with brain imaging, may help us shine a light in the black box that exists between human thinking, speech, behaviour and brain activity, between brain and mind.

But our problems helping people who suffer from schizophrenia are not really in the realm of diagnosis. And even if we had a more definitive way of making an early diagnosis of a particular type of schizophrenia those bigger problems remain.

And those problems lie in the realm of treatment, effective treatment without side effects, long term treatment, acceptance of the diagnosis, acceptance of the fact of having an illness, a mental illness, adequate resources, human resources and hospital beds and programs, compliance with treatment, public attitudes, and funding.

I doubt that an AI program, having been fed a recording of a patient’s conversation or speech pattern, and then announcing that this particular patient’s speech pattern shows a 76% likelihood of having schizophrenia, subtype 4, will do very much to alleviate the problems listed in the previous paragraph.

In the late 1970’s I saw a 13 year old girl with her parents. After a couple of meetings I referred this child to a colleague to help me with the diagnosis. He sent me an informal response after seeing her. It read: “David, this girl has schizophrenia. You know she has schizophrenia. You just don’t want to believe she has schizophrenia.”

About 1990 I was asked to intervene when a young male patient insisted on barging onto a different ward in the psychiatric hospital to see his sister. The young man had schizophrenia. The nurses assumed his delusions included having a non-existent sister. I intervened, and found he did have a sister on the ward: the 13 year old girl I had seen years before, now 27 years old and once again psychotic, and mentally and physically showing the ravages of time and chronic illness.

We don’t really need AI to help with diagnosis. But we could be spending money, both research and health care money, in other ways that would provide a better outcome for people who do suffer from schizophrenia.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *