Identification of voxel-based texture abnormalities as new biomarkers for schizophrenia and major depressive patients using layer-wise relevance propagation on deep learning decisions

Deep learning system for brain image-aided diagnosis of multiple major mental disorders

“…Research in other fields [56][57][58] has shown that deep learning is more accurate than traditional neuroimaging classification algorithms. However, the few current studies applying deep learning methods have used hand-crafted image features 39 or skipped active feature selection --treating the whole-brain image data equally 32 . Hand-crafted features may overwhelm fine-grained structural features of individuals due to spatial manipulation and averaging over regions.…”

“…Previous attempts have mainly focused on distinguishing certain types of diagnosis from health control 27,[29][30][31][32] . Some studies have built classifiers to distinguish between two or three diagnoses [33][34][35][36][37][38][39] , but the sensitivity and specificity of these classifiers have only been tested in two categories and not in the context of multiple diagnostic categories. Neuroimaging-assisted diagnostic systems that infer the possibility of multiple diagnostic types simultaneously and suggest possible co-morbid patterns are lacking.…”

Identification of texture MRI brain abnormalities on first-episode psychosis and clinical high-risk patients using explainable artificial intelligence

“…The de nitions of these features were as follows: GLCM-contrast re ects local variations in the GLCM; GLCM-energy re ects uniformity of greylevel voxel pairs; GLCM-entropy re ects randomness of grey-level voxel pairs; and GLCM-homogeneity re ects homogeneity of grey-level voxel pairs 57,58 ; nally GLCM-sum of entropy and difference of entropy re ect second order statistics of differentiation of grey level distribution GLCM. We used voxel-by-voxel sliding 3D cube of 7x7x7 dimension as presented in a previous paper 33 . The GLCM matrix was normalized by dividing the values with the total sum of the values in the matrix.…”

“…Radiomics texture features are able to quantify the hidden patterns between voxel intensities and the spatial distribution of these patterns across brain regions. Radiomics texture features with their potential as image-based biomarkers have been widely used across several studies, like for cancer identi cation 30 , Alzheimer's 31 and Parkinson's disease 32 as neurodegenerative diseases, major depression 33 and schizophrenia 34,35 . In the eld of schizophrenia research, texture features such as homogeneity and entropy have been shown to differentiate patients from healthy controls (HC) 36 .…”

Advances in Using MRI to Estimate the Risk of Future Outcomes in Mental Health - Are We Getting There?

“…The individual heatmaps corresponded to the lesions themselves and non-lesion gray and white matter areas such as the thalamus, which are conventional MRI markers ( 75 ). In a study where authors used texture feature maps for classifying participants with SZ, MD patients, and HC, LRP showed which zones contributed to the classification of the deep learning algorithm ( 94 ). Another interesting approach to determine which regions contribute the most to classification consists of substitute brain regions by healthy ones generated using variational autoencoders and then see how performance changes ( 95 ).…”