Neuroimaging-based classification of PTSD using data-driven computational approaches: A multisite big data study from the ENIGMA-PGC PTSD consortium.

BACKGROUND: Recent advances in data-driven computational approaches have been helpful in devising tools to objectively diagnose psychiatric disorders. However, current machine learning studies limited to small homogeneous samples, different methodologies, and different imaging collection protocols, limit the ability to directly compare and generalize their results. Here we aimed to classify individuals with PTSD versus controls and assess the generalizability using a large heterogeneous brain datasets from the ENIGMA-PGC PTSD Working group. METHODS: We analyzed brain MRI data from 3,477 structural-MRI; 2,495 resting state-fMRI; and 1,952 diffusion-MRI. First, we identified the brain features that best distinguish individuals with PTSD from controls using traditional machine learning methods. Second, we assessed the utility of the denoising variational autoencoder (DVAE) and evaluated its classification performance. Third, we assessed the generalizability and reproducibility of both models using leave-one-site-out cross-validation procedure for each modality. RESULTS: We found lower performance in classifying PTSD vs. controls with data from over 20 sites (60 % test AUC for s-MRI, 59 % for rs-fMRI and 56 % for d-MRI), as compared to other studies run on single-site data. The performance increased when classifying PTSD from HC without trauma history in each modality (75 % AUC). The classification performance remained intact when applying the DVAE framework, which reduced the number of features. Finally, we found that the DVAE framework achieved better generalization to unseen datasets compared with the traditional machine learning frameworks, albeit performance was slightly above chance. CONCLUSION: These results have the potential to provide a baseline classification performance for PTSD when using large scale neuroimaging datasets. Our findings show that the control group used can heavily affect classification performance. The DVAE framework provided better generalizability for the multi-site data. This may be more significant in clinical practice since the neuroimaging-based diagnostic DVAE classification models are much less site-specific, rendering them more generalizable.

Influence of analysis technique on measurement of diffusion tensor imaging parameters.

OBJECTIVE: We compared results from various methods of analysis of diffusion tensor imaging (DTI) data from a single dataset consisting of 10 healthy adolescents. SUBJECTS AND METHODS: All subjects were imaged on a single 3-T MRI system (single-shot echo-planar imaging pulse sequence; b value, 1000 s/mm(2)). We measured fractional anisotropy (FA), apparent diffusion coefficient (ADC), and axial and radial diffusivity values using 64-pixel rectangular regions of interest (ROIs) in the right side, midline, and left side of the central portion of the splenium of the corpus callosum for fixed (i.e., at same sites in all subjects) and targeted (i.e., at sites of highest FA values) locations. We compared results with those obtained using 64-pixel oval ROIs and 100-pixel rectangular ROIs in the same locations. Finally, we compared results from ROI-based methods and from tractography. All comparisons used the Wilcoxon signed rank test and the intraclass correlation of individual values. RESULTS: Compared to tractography, the average of mean ROI-based values was significantly higher for fixed (approximately 14%) and targeted (approximately 39%) FA values and was significantly lower for ADC (approximately 16%) and radial diffusivity (approximately 38%) values. For solely ROI-based comparisons, statistically significant differences were found in the following comparisons: 64- versus 100-pixel ROI, oval versus rectangular ROI, targeted FA left of midline versus mean targeted FA value, and targeted ROI right of midline versus mean targeted FA value. CONCLUSION: Markedly different values were obtained when using either ROI- or tractography-based techniques or ROI analysis techniques that differ only relatively slightly.

Diffusion-tensor imaging assessment of white matter maturation in childhood and adolescence.

OBJECTIVE: The purpose of this study was to test a first hypothesis that fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values continue to change in late childhood and adolescence and a second hypothesis that less mature white matter (WM) regions have a higher rate of change than WM regions that are relatively more mature. SUBJECTS AND METHODS: Eighty-seven healthy children (50 girls, 37 boys; mean age, 11.2 ± 3.6 years; range, 4.2-17.7 years) underwent six-direction diffusion-tensor imaging with a 3-T MRI system. Three neuroradiologists independently drew regions of interest in 10 WM regions and measured FA and ADC values. To test the first hypothesis, we correlated these values with subject age by linear regression analysis (p < 0.05). To test the second hypothesis, we determined whether regions with lower FA and higher ADC in the 4- to 7-year old group had a higher slope of FA increase and ADC decrease over the entire age range. For this assessment, we used linear regression analysis (p < 0.05) and curve fitting. RESULTS: In the test of the first hypothesis, increases in FA with age were noted in all WM regions and were statistically significant in six regions. Decreases in ADC values with age were noted in all brain regions except the genu of the corpus callosum. In all other regions except the splenium of the corpus callosum, the decreases were statistically significant. In the test of the second hypothesis, the relation between FA in the 4- to 7-year-old subjects and the FA increase in the entire sample was best described with a linear equation. The rate of age-related FA increase tended to be greater with lower initial FA (r = -0.384, p = 0.271). The relation between ADC in the 4- to 7-year-old subjects and ADC decrease in the entire population was best described with a second-order equation. The rate of age-related ADC decrease tended to be greater with higher initial ADC (r = 0.846, p = 0.001). For ADC values of 100 or less at age 4-7 years, the rate of ADC change with age tended to be decrease as initial ADC increased. CONCLUSION: In general, both hypotheses were verified. Overall, FA values continue to increase and ADC values continue to decrease during childhood and adolescence. The most rapid changes were found in WM regions that were least mature in the first few years of the study period.

Abnormalities in MRI-measured signal intensity in the corpus callosum in schizophrenia.

The microstructural integrity of the corpus callosum (CC) in first-episode schizophrenia patients was assessed by measuring the signal intensity (SI) in T1-weighted MRI images. Analyses revealed that compared to both healthy controls and non-schizophrenic patients, schizophrenia patients showed reductions in SI in all the callosal subregions, the genu, body, isthmus and splenium in first-episode schizophrenia. These results indicate that schizophrenia is characterized by pathology of this principal interhemispheric commissure; the abnormalities may reflect distributed (rather than localized) interhemispheric disconnectivity that extends beyond the heteromodal association cortices.

Cardiac implications of increased arterial entry and reversible 24-h central and peripheral norepinephrine levels in melancholia.

The mortality of chronic heart failure (CHF) doubles either when CHF patients are depressed or when their plasma norepinephrine (NE) level exceeds those of controls by approximately 40%. We hypothesized that patients with major depression had centrally driven, sustained, stress-related, and treatment-reversible increases in plasma NE capable of increasing mortality in CHF patients with depression. We studied 23 controls and 22 medication-free patients with melancholic depression. In severely depressed patients before and after electroconvulsive therapy (ECT), we measured cerebrospinal fluid (CSF) NE, plasma NE, plasma epinephrine (EPI), and plasma cortisol hourly for 30 h. In mildly-to-moderately depressed melancholic patients, we assessed basal and stress-mediated arterial NE appearance. Severely depressed patients had significant increases in mean around-the-clock levels of CSF NE (P < 0.02), plasma NE (P < 0.02), plasma EPI (P < 0.02), and plasma cortisol (P < 0.02). CSF NE, plasma NE, and cortisol all rose together throughout the night and peaked in the morning. Each fell to control values after ECT. Mildly-to-moderately melancholic patients also had increased basal (P < 0.05) and stress-related (P < 0.03) arterial NE-appearance rates. Severely melancholic depressed, medication-free patients had around-the-clock increases in plasma NE levels capable of increasing mortality in CHF. Twenty-four-hour indices of central noradrenergic, adrenomedullary, and adrenocortical secretion were also elevated. Concurrent diurnal rhythms of these secretions could potentiate their cardiotoxicity. Even mildly-to-moderately depressed melancholic patients had clinically relevant increases in the arterial NE-appearance rate. These findings will not apply to all clinical subtypes of major depression.