Application of Artificial Intelligence in the MRI Classification Task of Human Brain Neurological and Psychiatric Diseases: A Scoping Review

Abstract

Artificial intelligence (AI) for medical imaging is a technology with great potential. An in-depth understanding of the principles and applications of magnetic resonance imaging (MRI), machine learning (ML), and deep learning (DL) is fundamental for developing AI-based algorithms that can meet the requirements of clinical diagnosis and have excellent quality and efficiency. Moreover, a more comprehensive understanding of applications and opportunities would help to implement AI-based methods in an ethical and sustainable manner. This review first summarizes recent research advances in ML and DL techniques for classifying human brain magnetic resonance images. Then, the application of ML and DL methods to six typical neurological and psychiatric diseases is summarized, including Alzheimer’s disease (AD), Parkinson’s disease (PD), major depressive disorder (MDD), schizophrenia (SCZ), attention-deficit/hyperactivity disorder (ADHD), and autism spectrum disorder (ASD). Finally, the limitations of the existing research are discussed, and possible future research directions are proposed.

SPECT Functional Neuroimaging Distinguishes Adult Attention Deficit Hyperactivity Disorder From Healthy Controls in Big Data Imaging Cohorts

“…Machine learning or artificial intelligence (AI) techniques have been applied to neuroimaging in an effort to detect patterns and findings not evident from simple statistical analysis. Numerous AI techniques, such as support vector machine, multiple kernel learning, deep belief network, convolutional neural network and others have been applied to fMRI and anatomical MRI data ( 39 ). For example, a group of 36 adults with ADHD and 36 controls underwent anatomical MRI, fMRI using a cued attention task, and diffusion tensor imaging.…”

“…Notably, decreased activity of the right inferior frontal gyrus stood out as a strong predictor of ADHD status. The limitation of the AI work to date has been the relatively small sample sizes, the lack of consistent findings across studies ( 18 , 39 , 40 ) and the need for multiple time-consuming scans.…”

Sex classification from functional brain connectivity: Generalization to multiple datasets

“…Machine Learning (ML) is a powerful tool to relate neuroimaging data to behavior and phenotypes (Genon et al, 2022; Varoquaux & Thirion 2014) and is therefore increasingly being employed in neuroscience applications (Jollans et al, 2019; Buch et al, 2018; Varoquaux et al, 2018; Kohoutova et al 2020). Successful applications of ML approaches include the decoding of mental states (Haynes & Rees 2006), classification of mental disorders (Zhang et al 2021; Chen et al, 2020), as well as the prediction of demographic and behavioral phenotypes (Smith et al 2015; Nostro et al, 2018; Pläschke et al, 2020; Varikuti et al, 2018; More et al, 2023).…”

Machine learning for autism spectrum disorder diagnosis using structural magnetic resonance imaging: Promising but challenging

“…In ML, the key features are usually extracted manually and then tell the algorithm how to make a prediction or classification by consuming more information. For problems with complex nonlinear relationships, the DL algorithm is better suited because it learns features automatically and its performance is superior in image analysis fields, such as object detection and image classification ( Zhang et al, 2021 ). However, the diagnosis of ASD remains a formidable challenge, as studies based on ML have shown different results that may reflect the diversity of behavioral symptoms of the disorder and its proposed etiology, often linked to the brain ( Sivapalan and Aitchison, 2014 ).…”

“…Several publications ( Zhang L. et al, 2020 ; Zhang et al, 2021 ; Quaak et al, 2021 ) have reviewed the classification of ASD using only ML or DL algorithms. Some representative examples of previous reviews are listed in Table 1 .…”