Psychiatry HAS to lead AI adoption in Medicine
Artificial intelligence (AI) technology has the potential to revolutionize the field of psychiatry by improving the accuracy of diagnoses and therapeutic solutions, if only for its ability to observe, analyze and process huge amounts of data in a short time frame. Despite this potential, many psychiatrists have been slow to adopt AI technology, and may soon find themselves left behind, as the world of medicine, and the world in its entirety is adopting AI in unbelievable rate.
One challenge with implementing AI in psychiatry is the lack of trust in the technology among some clinicians. A study published in JAMA Network Open found that only 11% of psychiatrists were willing to use an AI algorithm to assist with diagnosis. The study authors suggested that this lack of trust may be due to concerns about the reliability and accuracy of AI algorithms.
Another challenge is the need for large amounts of high-quality data to train AI algorithms. In psychiatry, this can be particularly difficult because mental health data is often sensitive and protected by privacy laws. As a result, it can be challenging to collect and share enough data to develop robust AI algorithms.
Having said that, the challenges embeded in the DSM and ICD could be solved with ease by utilizing AI. Statistical classification of symptoms is currently the main system to determine a mental health disorder. While the human brain is a wonderful machine and meta-analysis, systematic reviews and manual collecitons were great for centuries, it is time to ramp up the way we collect, analyze and classify these disorders, and with the latest development of AI and ML (Should I really mention ChatGPT now? :) it is easier than ever.
There have been some notable successes in using AI in psychiatry. In one example, researchers at the University of California, San Francisco developed an AI algorithm that was able to accurately predict which patients with depression would benefit from cognitive behavioral therapy (CBT). The algorithm was able to achieve 80% accuracy, compared to 60% accuracy achieved by human clinicians.
Another example comes from a study published in The Lancet Digital Health, where researchers developed an AI algorithm that was able to accurately predict which patients with schizophrenia would respond positively to antipsychotic medication. The algorithm was able to predict response with 74% accuracy, compared to 61% accuracy achieved by human clinicians.
At Behavidence, following a full anonymization of our onboarding process, we have recruited more than 50,000 patients to train our AI models. Our AI algorithms have achieved +80% accuracy in detecting Depression, Anxiety, Stress and Post-partum depression. As far as we are aware, it is the most accurate and robustly trained set of psychiatric AI models of its kind.
We have started feeling the change in the industry with more and more clinicians areadopting our technology to assist with big-data-based observations that generates daily mental health score to achieve better health outcomes and reduce the cost of care.
It is encouraging to see the change, but Behavidence, its partners and competitors (Iβm looking at you Apple, Biogen, Healthyrhythms and co) have got to push this industry to completely revamp before it is too late. Can you imagine the DSM or ICD completely automated with classification models running in real time on real patients data?
We will share some groundbreaking news in this area very very soon!
Stay tuned.
