RESUMEN
OBJECTIVE: Electroencephalographic analysis (EEG) has emerged as a powerful tool for brain state interpretation. Studies have shown distinct deviances of patients with schizophrenia in EEG activation at specific frequency bands. METHODS: Evidence is presented for the validation of a Convolutional Neural Network (CNN) model using transfer learning for scalp EEGs of patients and controls during the performance of a speeded sensorimotor task and a working memory task. First, we trained a CNN on EEG data of 41 schizophrenia patients (SCZ) and 31 healthy controls (HC). Secondly, we used a pretrained model for training. Both models were tested in an external validation set of 15 SCZ, 16 HC, and 12 first-degree relatives. RESULTS: Using the layer-wise relevance propagation on the classification decision, a heatmap was produced for each subject, specifying the pixel-wise relevance. The CNN model resulted in the first case in a balanced accuracy of 63.7% and 81.5% in the second case, on the external validation test 64.5% and 83.2%, respectively. CONCLUSIONS: The theta and alpha frequency bands of the EEG signals had significant relevance to the CNN classification decision and predict the first-degree relatives indicating potential heritable functional deviances. SIGNIFICANCE: The proposed methodology results in important advancements for the identification of biomarkers in schizophrenia heritability.
Asunto(s)
Esquizofrenia , Encéfalo , Electroencefalografía/métodos , Humanos , Memoria a Corto Plazo , Redes Neurales de la Computación , Esquizofrenia/diagnósticoRESUMEN
Non-segmented MRI brain images are used for the identification of new Magnetic Resonance Imaging (MRI) biomarkers able to differentiate between schizophrenic patients (SCZ), major depressive patients (MD) and healthy controls (HC). Brain texture measures such as entropy and contrast, capturing the neighboring variation of MRI voxel intensities, were computed and fed into deep learning technique for group classification. Layer-wise relevance was applied for the localization of the classification results. Texture feature map of non-segmented brain MRI scans were extracted from 141 SCZ, 103 MD and 238 HC. The gray level co-occurrence matrix (GLCM) was calculated on a voxel-by-voxel basis in a cube of voxels. Deep learning tested if texture feature map could predict diagnostic group membership of three classes under a binary classification (SCZ vs. HC, MD vs. HC, SCZ vs. MD). The method was applied in a repeated nested cross-validation scheme and cross-validated feature selection. The regions with the highest relevance (positive/negative) are presented. The method was applied on non-segmented images reducing the computation complexity and the error associated with segmentation process.