Study finds cross-species machine learning improves accuracy of magnetic resonance imaging diagnosis of mental illness

On 17 June, the American Journal of The Journal of The Journal of The Journal published an online research paper entitled Diagnostic Classification for Human Autism and The-Copulsive Disorder on Machine Learning From from a Primate Genetic Modelthe research was completed by wang Zheng Research Group of the Center for Excellence in Brain Science and Intelligent Technology of the Chinese Academy of Sciences (Institute of Neuroscience), Shanghai Brain Science and Brain Research Center, And The National Key Laboratory of Neuroscience, in collaboration with the Research Group of researcher Shiman, Institute of Automation of the Chinese Academy of Sciences, to integrate primate models and functional magnetic resonance imaging data for patients with clinical mental illnessThe process of machine learning analysis, using the characteristics learned from the genetically modified macaque model, constructs the classifier model of clinical mental illness patients, then analyzes the neural loop mechanism of human autism and obsessive compulsive disorder in depth, provides new evidence for the accuratediagnosisof mental illness, and discovers a new way to use non-human primate models to serve the clinical application needs of mental illnessautism (ASD) is a developmental disorder of the nervous system, with a high degree of heterogeneity, and patients often accompany complications such as obsessive compulsive disorder (OCD), attention deficit hyperactivity disorder (ADHD), which poses great challenges to clinicaldiagnosisand pathological mechanismsNon-human primate model animals and humans are closer to brain structure and function, the researchers previously found that genetically modified primate models can show similar symptompoids to human clinical patients, such as MECP2 over-expression of macaques exhibiting repeated stereotyped behavior, social behavior disorders and other autistic symptoms (Nature, 2016), and abnormalities in the brain loop are similar to some autistic patients (Jsci, 2020)the team of researchers based on the previous work (IEEE TMI, 2015) to explore possible evolutionary conservative characteristics among primate species, assuming that a predictive model of classification of mental illness that can migrate across species is based on conservative brain region function (Figure A)The study used Group LASSO algorithms to screen brain area data from brain function maps from five genetically modified macaques
and 11 wild macaques to identify nine core brain regions (Figure B); Patient data came from 4 clinical imaging databases, such as ABIDE-I (1112), ABIDE-II (1114), OCD (186), and ADHD-200 (776)After cross-validation, the study found that the classification model based on the characteristics of genetically modified macaques had an accuracy of 82.14% for the distinction between autistic and normal people in the ABIDE-I data set, and 75.17% of the human subjects in the ABIDE-II database, significantly higher than the performance of building classifiers based on the characteristics of autism and obsessive-compulsive disorder patients (Figure C)When the same nine brain regions were extended to obsessive compulsive imaging data, the study found that macaque features still achieved 78.36 percent accuracy, significantly higher than classifier performance based on the characteristics of autistic peopleHowever, these characteristics of monkey-based model learning failed to significantly improve the accuracy of classification in patients with ADHDFurther analysis of the relationship between functional connections in these superior classifiers and clinical symptoms of mental illness was found to show that the outer exfrontal cortex in the right abdomen played a dual role in both autism and obsessive compulsive disorder, corresponding to their specific dimensional symptom phenotypes (Figure D)(
BiovalleyBioon.com)