Objective depression diagnosis based on brain networks — AI-assisted medical device approved by Ministry of Health, Labour and Welfare

While psychiatric and neurological disorders are considered to be disorders of the brain network, diagnosis is based solely on symptoms, and currently only 4% of diagnostic criteria are said to be reliable. Establishing objective diagnosis and restructuring diagnosis based on brain networks has been a long-standing goal of psychiatric medicine. XNef (Soraku District, Kyoto Prefecture), the Advanced Telecommunications Research Institute International (ATR, Soraku District, Kyoto Prefecture), and the Department of Psychiatry and Neurosciences, Graduate School of Biomedical and Health Sciences, Hiroshima University have jointly developed a program medical device that assists in the diagnosis of depression based on fMRI data, and have obtained approval from the Ministry of Health, Labour and Welfare.

Director Mitsuo Kawato of the Brain Information Communication Research Laboratory Group at ATR (Chief Executive Officer of XNef), stated: "From the Strategic Research Program for Brain Sciences to the current Brain and Neural Science Integration Program, we have advanced projects with the slogan 'brain science that contributes to society,' and we have finally reached the stage of obtaining approval from the Pharmaceuticals and Medical Devices Agency (PMDA). We will continue to accumulate clinical evidence and advance stratified medicine for depression, as well as apply this to other psychiatric and neurological disorders."

Depression is a syndrome consisting of a collection of various symptoms including depressed mood, loss of interest and pleasure, weight (appetite) changes, insomnia or hypersomnia, psychomotor agitation or retardation, fatigue or loss of energy, feelings of worthlessness or guilt, and decreased thinking or concentration abilities. In clinical psychiatry, diagnosis is made after confirming symptoms through interviews and conducting MRI examinations to exclude organic causes (such as brain tumors or stroke), and treatment methods are determined. In other words, there are no biological markers for diagnosis.

Even experienced psychiatrists experience variability in diagnosis. More than 90% of depression patients visit general practitioners such as internal medicine doctors, but their diagnostic accuracy is about 50%. To evaluate psychiatric and neurological disorders from the state of the brain, their networks must be evaluated. For this purpose, resting-state fMRI, which can evaluate brain networks through fluctuations in resting brain activity, was used. This can be acquired during about 10 minutes of imaging during MRI examination for excluding organic causes and can evaluate individual brain networks.

By utilizing AI to analyze resting-state fMRI data from healthy groups and depression groups, depression brain network markers can be created. While such attempts have been made numerous times, there was the issue that markers created at one facility could not be applied to other facilities due to AI overfitting.

From this, by utilizing part of the resting-state fMRI data from 4,625 cases (2,031 healthy subjects, 944 depression patients, 233 bipolar disorder patients, etc.) from 11 domestic facilities, a generalizable brain network marker was constructed. When verified with datasets from 285 healthy subjects and 236 depression patients from five hospitals, an accuracy rate of 70% was achieved, succeeding in discrimination at different and completely independent facilities for the first time in the world.

In this study, prospective verification was conducted with MRI from multiple facilities and multiple vendors. Since stability more than twice that of physician diagnosis could be secured, this brain network marker obtained manufacturing and sales approval for the first time as an AI-assisted program medical device within the framework of two-stage approval. As a specific method of use, resting-state fMRI image data of patients visiting hospitals is automatically anonymized and sent to a dedicated server, and physicians can use it for diagnosis by confirming the analysis results.

Currently, Hiroshima University is conducting specific clinical trials at 8 medical facilities in the city and clinical trials at other sites with the Institute of Science Tokyo as the representative medical institution. Based on the results of these trials, clinical evidence will be accumulated to obtain second-stage approval.

The research group is also advancing verification of depression stratification. Currently, about 20% of depression cases achieve remission in 8 weeks, while 65% take 3.5 years to achieve remission.

Principal Researcher Yuki Sakai at ATR (Vice President of XNef), stated: "Since depression is a syndrome involving various combinations of symptoms, it is not easy to choose which treatment will be most effective from the beginning." Based on brain network patterns, depression patients are classified into biotypes: those responsive to Drug A, those responsive to Drug B, those for whom cognitive behavioral therapy is effective, and those for whom repetitive transcranial magnetic stimulation therapy is effective. "We have confirmed that treatment responsiveness differs for each subtype, and in the verification stage, we have prospects for about 50% remission in 8 weeks and 80% remission in 16 weeks" (Sakai). They plan to proceed with approval applications in the future.

Kawato stated: "Currently, depression is the target, but by exploring biological backgrounds, we want to connect this to the diagnosis and treatment of other psychiatric disorders."

Journal Information
Publication: Journal of Affective Disorders
Title: Verification of the brain network marker of major depressive disorder: Test-retest reliability and anterograde generalization performance for newly acquired data
DOI: 10.1016/j.jad.2023.01.087