Deep convolutional neural networks (DCNNs) have achieved great success for image classification in medical research. Deep learning with brain imaging is the imaging method of choice for the diagnosis and prediction of Alzheimer’s disease (AD). However, it is also well known that DCNNs are “black boxes” owing to their low interpretability to humans. The lack of transparency of deep learning compromises its application to the prediction and mechanism investigation in AD. To overcome this limitation, we develop a novel general framework that integrates deep leaning, feature selection, causal inference, and genetic-imaging data analysis for predicting and understanding AD. The proposed algorithm not only improves the prediction accuracy but also identifies the brain regions underlying the development of AD and causal paths from genetic variants to AD via image mediation. The proposed algorithm is applied to the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset with diffusion tensor imaging (DTI) in 151 subjects (51 AD and 100 non-AD) who were measured at four time points of baseline, 6 months, 12 months, and 24 months. The algorithm identified brain regions underlying AD consisting of the temporal lobes (including the hippocampus) and the ventricular system.
Background: The use of magnetic resonance imaging (MRI) in diagnosis of neonatal acute bilirubin encephalopathy (ABE) in newborns has been limited by its difficulty in differentiating confounding image contrast changes associated with normal myelination. This study aims to demonstrate the feasibility of building a machine learning prediction model based on radiomics features derived from MRI to better characterize and distinguish ABE from normal myelination.
Methods: In this retrospective study, we included 32 neonates with clinically confirmed ABE and 29 age-matched controls with normal myelination. Radiomics features were extracted from the manually segmented region of interest (ROI) on T1-weighted spin echo images, followed by the feature selection using two-sample independent t-test, least absolute shrinkage and selection operator (Lasso) regression, and Pearson's correlation matrix. Additional feature quantifying the relative mean intensity of ROI was defined and calculated. A prediction model based on the selected features was built to classify ABE and normal myelination using multiple machine learning classifiers and a leave-one-out cross-validation scheme. Receiver operating characteristics (ROC) analysis was used to evaluate the prediction performance with the area under the curve (AUC) and feature importance ranked based on the Fisher score.
Results: Among 1319 radiomics features, one radiologist-defined intensity-based feature and 12 texture features were selected as the most discriminative features. Based on these features, decision trees had the best classification performance with the largest AUC of 0.946, followed by support vector machine (SVM), tree-bagger, logistic regression, Naïve Bayes, discriminant analysis, and k-nearest neighborhood (KNN), which have an AUC of 0.931, 0.925, 0.905, 0.891, 0.883, and 0.817, respectively. The relative mean intensity outperformed other 12 texture features in differentiating ABE from controls.
Conclusions: The results from this study demonstrated a new strategy of characterizing ABE-induced intensity and morphological changes in MRI, which are difficult to be recognized, interpreted, or quantified by the routine experience and visual-based reading strategy. With more quantitative and objective measurements, the reported machine learning assisted radiomics features-based approach can improve the diagnosis and support clinical decision-making.
Objective: The anterior cingulate cortex (ACC) is associated with the processing of negative emotions. Gamma-aminobutyric acid (GABA) metabolism plays an important role in the pathogenesis of mental disorders. We aimed to determine the changes in GABA levels in the ACC of perimenopausal women with depression.
Methods: We recruited 120 perimenopausal women, who were followed up for 18–24 months. After reaching menopause, the participants were divided into a control group (n = 71), an anxiety group (n = 30), and a depression group (n = 19). The participants were examined using proton magnetic resonance spectroscopy (MRS). TARQUIN software was used to calculate the GABA concentrations in the ACC before and after menopause. The relationship of the GABA levels with the patients’ scores on the 14-item Hamilton Anxiety Scale and 17-item Hamilton Depression Scale was determined.
Results: GABA decreased with time. The postmenopausal GABA levels were significantly lower in the depression group than in the anxiety group and were significantly lower in both these groups than in the normal group. The postmenopausal GABA levels were significantly lower than the premenopausal levels in the normal, anxiety, and depression groups (P = 0.014, <0.001, and <0.001, respectively). The premenopausal GABA levels did not significantly differ between the normal vs. anxiety group (P = 0.907), normal vs. depression group (P = 0.495), and anxiety vs. depression group. The postmenopausal GABA levels were significantly lower in the depression group than in the anxiety group and were significantly lower in both these groups than in the normal group, normal vs. anxiety group (P = 0.022), normal vs. depression group (P < 0.001), and anxiety vs. depression group (P = 0.047).
Conclusion: Changes in GABA concentrations in the anterior cingulate cortex are related with the pathophysiological mechanism and symptoms of perimenopausal depression.
Objectives: To evaluate white matter hyperintensities (WMH) quantification reproducibility from multiple aspects of view and examine the effects of scan–rescan procedure, types of scanner, imaging protocols, scanner software upgrade, and automatic segmentation tools on WMH quantification results using magnetic resonance imaging (MRI).
Methods: Six post-stroke subjects (4 males; mean age = 62.8, range = 58–72 years) were scanned and rescanned with both 3D T1-weighted, 2D and 3D T2-weighted fluid-attenuated inversion recovery (T2-FLAIR) MRI across four different MRI scanners within 12 h. Two automated WMH segmentation and quantification tools were used to measure WMH volume based on each MR scan. Robustness was assessed using the coefficient of variation (CV), Dice similarity coefficient (DSC), and intra-class correlation (ICC).
Results: Experimental results show that the best reproducibility was achieved by using 3D T2-FLAIR MRI under intra-scanner setting with CV ranging from 2.69 to 2.97%, while the largest variability resulted from comparing WMH volumes measured based on 2D T2-FLAIR MRI with those of 3D T2-FLAIR MRI, with CV values in the range of 15.62%–29.33%. The WMH quantification variability based on 2D MRIs is larger than 3D MRIs due to their large slice thickness. The DSC of WMH segmentation labels between intra-scanner MRIs ranges from 0.63 to 0.77, while that for inter-scanner MRIs is in the range of 0.63–0.65. In addition to image acquisition, the choice of automatic WMH segmentation tool also has a large impact on WMH quantification.
Conclusion: WMH reproducibility is one of the primary issues to be considered in multicenter and longitudinal studies. The study provides solid guidance in assisting multicenter and longitudinal study design to achieve meaningful results with enough power.
KEY POINTS
- The intra-scanner and inter-scanner WMH reproducibility study in the same cohort.
- The best reproducibility was achieved by using 3D T2-FLAIR MRI under intra-scanner setting.
- There is a large variability in comparing WMH quantification results based on 2D T2-FLAIR MRI with those of 3D T2-FLAIR MRI.
Background: Analyses of resting-state functional magnetic resonance imaging (rs-fMRI) have been performed to investigate pathophysiological changes in the brains of patients with autism spectrum disorder (ASD) relative to typically developing controls (CTLs). However, the results of these previous studies, which have reported mixed patterns of hypo- and hyperconnectivity, are controversial, likely due to the small sample sizes and limited age range of included participants.
Methods: To overcome this issue, we analyzed multisite neuroimaging data from a large sample (n = 626) of male participants aged between 5 and 29 years (mean age = 13 years). The rs-fMRI data were preprocessed using SPM12 and DPARSF software, and signal changes in 90 brain regions were extracted. Multiple linear regression was used to exclude the effect of site differences in connectivity data. Subcortical–cortical connectivity was computed using connectivities in the hippocampus, amygdala, caudate nucleus, putamen, pallidum, and thalamus. Eighty-eight connectivities in each structure were compared between patients with ASD and CTLs using multiple linear regression with group, age, and age × group interactions, head movement parameters, and overall connectivity as variables.
Results: After correcting for multiple comparisons, patients in the ASD group exhibited significant increases in connectivity between the thalamus and 19 cortical regions distributed throughout the fronto-parietal lobes, including the temporo-parietal junction and posterior cingulate cortices. In addition, there were significant decreases in connectivity between the amygdala and six cortical regions. The mean effect size of hyperconnectivity (0.25) was greater than that for hypoconnectivity (0.08). No other subcortical structures showed significant group differences. A group-by-age interaction was observed for connectivity between the thalamus and motor-somatosensory areas.
Conclusions: These results demonstrate that pathophysiological changes associated with ASD are more likely related to thalamocortical hyperconnectivity than to amygdala-cortical hypoconnectivity. Future studies should examine full sets of clinical and behavioral symptoms in combination with functional connectivity to explore possible biomarkers for ASD.
The notion of dysconnectivity in schizophrenia has been put forward for many years and results in substantial attempts to explore altered functional connectivity (FC) within different networks with inconsistent results. Clinical, demographical, and methodological heterogeneity may contribute to the inconsistency. Forty-four patients with first-episode, drug-naive schizophrenia, 42 unaffected siblings of schizophrenia patients and 44 healthy controls took part in this study. Global-brain FC (GFC) was employed to analyze the imaging data. Compared with healthy controls, patients with schizophrenia and unaffected siblings shared enhanced GFC in the left superior frontal gyrus (SFG). In addition, patients had increased GFC mainly in the thalamo-cortical network, including the bilateral thalamus, bilateral posterior cingulate cortex (PCC)/precuneus, left superior medial prefrontal cortex (MPFC), right angular gyrus, and right SFG/middle frontal gyrus and decreased GFC in the left ITG/cerebellum Crus I. No other altered GFC values were observed in the siblings group relative to the control group. Further ROC analysis showed that increased GFC in the left SFG could separate the patients or the siblings from the controls with acceptable sensitivities. Our findings suggest that increased GFC in the left SFG may serve as a potential endophenotype for schizophrenia.
Background: Schizophrenia is characterized by the disruption of microstructural white matter (WM) integrity, while the pathogenesis remains unclear. Inflammation has been associated with the WM pathology in schizophrenia. Interleukin 10 (IL-10) has been proven to be related to schizophrenia in both animal and human models. The aim of this study was to explore whether peripheral IL-10 was associated with microstructural WM integrity in schizophrenia.
Methods: A total of 47 patients with schizophrenia (SZ) and 49 healthy controls (HC) underwent diffusion tensor imaging and venous blood sampling. Tract-based spatial statistics was conducted to explore the differences in fractional anisotropy (FA), radial diffusivity (RD), mean diffusivity (MD), and axial diffusivity (AD) between patients and controls. A quantitative chemiluminescence assay was performed to measure peripheral IL-10 levels. General linear regression analysis using a stepwise method was applied to examine the relationship between peripheral IL-10 and diffusion measures.
Results: Compared with the HC, peripheral IL-10 levels were higher and a significant reduction of FA and AD, and increase of RD and MD were observed in SZ (corrected p < 0.05). A regression analysis revealed that peripheral IL-10 was negatively correlated with FA in the right posterior thalamic radiation and left inferior fronto-occipital fasciculus, in SZ (β = -0.51, p = 0.01; β = -0.47, p = 0.02, respectively) but not in HC (β = -0.01, p = 0.95; β = -0.003, p = 0.98, respectively), and the differences in regression curves were significant (z = 2.50, p = 0.01; z = 2.37, p = 0.02, respectively). IL-10 was negatively connected with MD in the right parietal arcuate fasciculus (β = -0.40, p = 0.048) and body of the corpus callosum (β = -0.43, p = 0.03) in SZ, while not in HC. The magnitude of correlation in the patient and control group was different (z = 2.48, p = 0.01 and z = 2.61, p < 0.01, respectively). In addition, IL-10 was positively correlated with RD in the right parietal arcuate fasciculus in patients (β = 0.45, p = 0.04) but not in HC (β = 0.26, p = 0.94), but the correlation coefficients were not significant (z = 0.98, p = 0.32).
Conclusion: Our findings demonstrated that elevated peripheral IL-10 levels were associated with the disruption of microstructural WM integrity in schizophrenia, supporting the notion that inflammation plays a regulatory role in the pathology of microstructural WM and is associated with schizophrenia.