Deep Learning System for Brain Image-Aided Diagnosis of Multiple Mental Disorders

Clinical diagnoses of psychiatric disorders rely on clinical symptom assessments. While neuroimaging has been applied to characterize the brain in psychiatric disorders, it has not been used to diagnose major psychiatric disorders. Here, we report a neuroimaging-aided diagnostic system for major psychiatric disorders designed for clinical practice. We developed deep learning networks with attention mechanisms and applied them to a single-center neuroimaging dataset containing data from four psychiatric disorders and healthy controls (n = 2490). Cross-validations and independent validations using multicenter (n = 1346) and real-world datasets (n = 344) showed that the system used brain images to identify bipolar disorder, major depressive disorder, schizophrenia, and obsessive-compulsive disorder. The classifiers also reflected comorbidity diagnoses. Furthermore, we implemented an automated closed-loop system that provides guidance in real-world clinical settings and that further learns by accumulating data and feedback. This work advances brain research in its aim to provide objective diagnostic aid for identifying psychiatric disorders for real-world use.

Note:
Funding Information: This work was supported by National Natural Science Foundation of China (81971682, 81571756, 81270023); Natural Science Foundation of Shanghai (20ZR1472800); Shanghai Municipal Commission of Health (2018BR17, 2019ZB0201); Shanghai Clinical Research Center for Mental Health (19MC1911100); Shanghai Municipal Commission of Education- Gaofeng Clinical Medicine Grant Support (20171929); Shanghai Science and Technology Commission (18JC1420305); Shanghai Intelligent Engineering Technology Research Center for Addiction and rehabilitation (19DZ2255200); Shanghai Municipal Health Bureau (201840269); Clinical Research Center of Shanghai Mental Health Center (CRC2018DSJ01- 5; CRC2019ZD04); Research Funds from Shanghai Mental Health Center (13dz2260500, 2018-YJ-02).

Conflict of Interests: ZY, QL, YH, and XZ hold the following patents related to this work: - A system that enables neuroimaging-assisted diagnostic processing for psychiatric disorders. China Patent 202111212750.1 - Method, system, apparatus, processor and storage medium for normalizing brain image features based on healthy population distribution. China Patent 202111208007.9 - Method, system, apparatus, processor, and storage medium for adaptive segmentation of medical image domains based on single-center calibration data. China Patent 202110757216.2. All authors declare no competing interests with manuscript.

Ethical Approval: The IRB committee of the Shanghai Mental Health Center approved this study and permitted the use of measures extracted from the coded data (IRB2020-15).

Keywords: Deep Learning, image-aided diagnosis, attention, mental disorders

Suggested Citation: Suggested Citation

Li, Qingfeng and Wang, Wenzheng and Guo, Qin and Jiang, Lijuan and Qiao, Kaini and Hu, Yang and Zhang, Xiaochen and Wang, Zhen and Peng, Daihui and Fan, Qing and Zhao, Min and Fang, Yiru and Wang, Jijun and Qiu, Hong and Wang, Jinhong and Li, Guanjun and Sheng, Jianhua and Jin, Cheng and Shen, Dinggang and Niznikiewicz, Margaret and Li, Chunbo and Consortium, The Psychiatric Imaging and Yang, Zhi and Administrator, Sneak Peek, Deep Learning System for Brain Image-Aided Diagnosis of Multiple Mental Disorders. Available at SSRN: https://ssrn.com/abstract=4496774 or http://dx.doi.org/10.2139/ssrn.4496774

This version of the paper has not been formally peer reviewed.