Objective Currently, depression diagnosis relies primarily on behavioral symptoms and signs, and treatment is guided by trial and error instead of evaluating associated underlying brain characteristics. Unlike past studies, we attempted to estimate accurate prediction models for late-life depression diagnosis and treatment response using multiple machine learning methods with inputs of multi-modal imaging and non-imaging whole brain and network-based features. Methods Late-life depression patients (medicated post-recruitment) [n=33] and elderly non-depressed individuals [n=35] were recruited. Their demographics and cognitive ability scores were recorded, and brain characteristics were acquired using multi-modal magnetic resonance imaging pre-treatment. Linear and nonlinear learning methods were tested for estimating accurate prediction models. Results A learning method called alternating decision trees estimated the most accurate prediction models for late-life depression diagnosis (87.27% accuracy) and treatment response (89.47% accuracy). The diagnosis model included measures of age, mini-mental state examination score, and structural imaging (e.g. whole brain atrophy and global white mater hyperintensity burden). The treatment response model included measures of structural and functional connectivity. Conclusions Combinations of multi-modal imaging and/or non-imaging measures may help better predict late-life depression diagnosis and treatment response. As a preliminary observation, we speculate the results may also suggest that different underlying brain characteristics defined by multi-modal imaging measures—rather than region-based differences—are associated with depression versus depression recovery since to our knowledge this is the first depression study to accurately predict both using the same approach. These findings may help better understand late-life depression and identify preliminary steps towards personalized late-life depression treatment.

Indices of functional connectivity in the default mode network (DMN) are promising neural markers of treatment response in late-life depression. We examined the differences in DMN functional connectivity between treatment-responsive and treatment-resistant depressed older adults. Forty-seven depressed older adults underwent MRI scanning pre- and post- pharmacotherapy. Forty-six never depressed older adults underwent MR scanning as comparison subjects. Treatment response was defined as achieving a Hamilton Depression Rating Scale of 10 or less post-treatment. We analyzed resting state functional connectivity using the posterior cingulate cortex as the seed region-of-interest. The resulting correlation maps were employed to investigate between-group differences. Additionally we examined the association between white matter hyperintensity burden and functional connectivity results. Comparison of pre- and post-treatment scans of depressed participants revealed greater post-treatment functional connectivity in the frontal precentral gyrus. Relative to treatment-responsive participants, treatment-resistant participants had increased functional connectivity in the left striatum. When adjusting for white matter hyperintensity burden, the observed differences lost significance for the PCC-prefrontal functional connectivity, but not for the PCC-striatum functional connectivity. The post-treatment “frontalization” of the DMN connectivity suggests a normalizing effect of antidepressant treatment. Moreover, our study confirms the central role of white matter lesions in disrupting brain functional connectivity.

Objective This study tests whether or not the structural white matter lesions that are characteristic of late-life depression are associated with alterations in the functional affective circuits of late-life depression. This study used an emotional faces paradigm that has been shown to engage the affective limbic brain regions. Method Thirty-three elderly depressed patients and 27 nondepressed comparison subjects participated in this study. The patients were recruited through the NIMH-sponsored Advanced Center for Interventions and Services Research for the Study of Late-Life Mood Disorders at the University of Pittsburgh Center for Bioethics and Health Law. Structural and functional MRI was used to assess white matter hyperintensity (WMH) burden and functional magnetic resonance imaging (fMRI) blood-oxygen-level-dependent (BOLD) response on a facial expression affective-reactivity task in both elderly participants with nonpsychotic and non-bipolar major depression (unmedicated) and nondepressed elderly comparison subjects. Results As expected, greater subgenual cingulate activity was observed in the depressed patients relative to the nondepressed comparison subjects. This same region showed greater task-related activity associated with a greater burden of cerebrovascular white matter change in the depressed group. Moreover, the depressed group showed a significantly greater interaction of WMH by fMRI activity effect than the nondepressed group. Conclusions The observation that high WMH burden in late-life depression is associated with greater BOLD response on the affective-reactivity task supports the model that white matter ischemia in elderly depressed patients disrupts brain mechanisms of affective regulation and leads to limbic hyperactivation.

Depression is a complex clinical entity that can pose challenges for clinicians regarding both accurate diagnosis and effective timely treatment. These challenges have prompted the development of multiple machine learning methods to help improve the management of this disease. These methods utilize anatomical and physiological data acquired from neuroimaging to create models that can identify depressed patients vs. non-depressed patients and predict treatment outcomes. This article (1) presents a background on depression, imaging, and machine learning methodologies; (2) reviews methodologies of past studies that have used imaging and machine learning to study depression; and (3) suggests directions for future depression-related studies.

White matter hyperintensities (WMHs) are often identified on T2-weighted magnetic resonance (MR) images in the elderly. The WMHs are generally associated with small vessel ischemic or pre-ischemic changes. However, the association of WMHs with blood oxygen level dependent (BOLD) functional magnetic resonance imaging (fMRI) signal is understudied. In this study, we evaluate how the BOLD signal change is related to the presence of WMHs in the elderly. Data were acquired as part of a study of late-life depression and included elderly individuals with and without major depression. The subjects were pooled because the presence of depression was not significantly associated with task-related BOLD changes, task performance, and WMH distribution. A whole brain voxel-wise regression analysis revealed a significant negative correlation between WMH burden and BOLD signal change during finger-tapping in the parietal white matter. Our observation that WMHs are associated with a significant diminution of the BOLD signal change underscores the importance of considering cerebrovascular burden when interpreting fMRI studies in the elderly. The mechanism underlying the association of WMH and BOLD signal change remains unclear: the association may be mediated by changes in neural activation, changes in coupling between neuronal activity and hemodynamics, or, perhaps, secondary to the effect of the ischemic changes on the sensitivity of the T2* BOLD MR signal.