Automated vessel segmentation in brain CT angiography (CTA) remains challenging despite the potential benefits of its applications. Expert acquisition of reference vessel segmentations is a laborious task. We propose an unsupervised generative deep learning approach that can be trained for vessel segmentation in brain CTA using a large dataset (n=908) of unlabelled brain CTAs and non-contrast enhanced CTs (NCCTs). Our semi-supervised approach uses a conditional generative adversarial network (GAN) for CTA to NCCT translation by generating a contrast map that allows for automatic extraction of vessel segmentations. Furthermore, we propose a 3D Frangi filter-based loss function to enhance tubular structures in the contrast map to improve vessel segmentations. We used a hold-out test set of 9 CTA volumes with manually annotated reference segmentations. We compared our semi-supervised approach with a state-of-the-art supervised nnUnet, trained and evaluated with test set using 9-fold nested cross-validation. Evaluation metrics included voxel-wise Dice similarity coefficient (DSC), true positive rate (TPR), and false positive rate (FPR). The DSC was 4 % lower for the semi-supervised approach (DSC: 0.74) compared to the supervised nnUnet (DSC: 0.78). Both the TPR and FPR were higher for the semi-supervised approach (TPR: 0.75, FPR/1000 voxels:2.05) compared to the supervised nnUnet (TPR:0.71, FPR/1000 voxels:0.87). Hence, the quantitative results showed that our semi-supervised method approaches a supervised state-of-the-art segmentation network. The results demonstrate that a semi-supervised generative deep learning approach for the segmentation of intracranial vessels is feasible without laborious manual segmentations.

AIMS:Major depressive disorder (MDD) is widely considered to be a mood disorder characterized by altered connectivity. The thalamus plays an important role in MDD by connecting large areas of the brain. Here, we explored thalamic connectivity in MDD at the subnuclear level using ultra-high-field MRI. METHODS:We combined ultra-high-field functional and diffusion MRI at 7.0 Tesla to map the connectivity of thalamic subnuclei in MDD patients (n = 47) and healthy controls (n = 13). We segmented thalamic subnuclei and calculated the functional and structural connectivity of thalamic subnuclei and tested for group differences and associations with clinical characteristics. RESULTS:MDD patients showed increased functional connectivity of the right thalamic central lateral nucleus with the right amygdala, bilateral inferior occipital lobe, and right transverse temporal gyrus, which was accompanied by increased structural connectivity between the right thalamic central lateral nucleus and the right inferior occipital lobe. Medicated MDD patients had a greater streamline count of the right thalamic central lateral nucleus-right inferior occipital lobe tract than healthy controls. CONCLUSIONS:Thalamic subnuclear connectivity with cortical and subcortical brain regions is perturbed in MDD. These results further support increased thalamic connectivity in MDD and suggest that this is related to a specific subnucleus.

This study aimed to obtain high-resolution 3D isotropic turbo spin-echo (TSE) wrist MRI acquisitions at 7T, with and without fat suppression, facilitated by compressed-sensing (CS) acceleration. In 16 healthy subjects, fat-suppressed (FS) and nonfat-suppressed (NFS) TSE wrist images were obtained. The protocol consisted of a SENSE-accelerated scan, with an isotropic voxel size of 0.45 mm and acquisition time of 7 min ("SENSE45"), a 0.45-mm, 4-min CS-accelerated scan ("CS45"), and a 0.35-mm, 7-min CS-accelerated scan ("CS35"). For two subjects, additional 0.45-mm, 4-min SENSE-accelerated scans were acquired ("High-SENSE"). For the NFS scans, refocusing pulses were optimized to mitigate water-fat chemical-shift artifacts in the slab-selection direction. Anatomical visibility of wrist structures and image quality were assessed qualitatively and through musculoskeletal radiologist grading. The use of nonselective hard refocusing pulses with optimized bandwidths and a center frequency in between water and fat enabled NFS imaging. The image quality of the faster CS45 scans was lower than for SENSE45, with statistically significantly different grading in 9/16 (FS) and 2/6 (NFS) grading parameters. Nonetheless, a similar scan time reduction could not be achieved using High-SENSE. No distinct benefit of CS35 compared to SENSE45 was evident in either the FS or NFS scans. NFS CS35 exhibited enhanced bone sharpness compared to SENSE45 for some subjects, yet on a group level, the difference was not statistically significant. In conclusion, for maintained voxel size, CS presents the opportunity to achieve shorter scan times than possible with SENSE alone, but with reduced image quality. For maintained scan time, although higher resolution CS incidentally showed a promising increase in NFS bone sharpness compared to SENSE, it does not present an unequivocal advantage for 3D 7-T TSE wrist MRI at this stage. Further optimization of the acquisition and reconstruction process is recommended.

Ultra-high field MRI facilitates imaging at high spatial resolutions, which may become important for detailed anatomical and pathological assessment of the human liver. Therefore, we aimed to advance structural liver imaging at 7 T by implementing a high-resolution, phase-shimmed, free-breathing liver scan. Six healthy participants underwent liver MRI scans at 7 T, utilizing an eight-channel parallel transmission system for phase shimming. B0 mapping and Fourier phase-encoded dual refocusing echo acquisition mode (PE-DREAM) multichannel B1 + mapping were performed during breath-holds at expiration. Prospectively undersampled golden-angle pseudo-spiral k-space data were acquired under free breathing, enabling retrospective respiratory binning using self-gating. Post-binning, the simultaneous autocalibrating and k-space estimation (SAKE) algorithm was employed for interpolation of a center of k-space area, prior to estimation of receive coil sensitivity maps. Image reconstruction was performed on expiration-phase data using compressed sensing, optimizing image quality by evaluating various regularization factors and numbers of respiratory bins. Finally, N4BiasFieldCorrection was applied to the resulting images. Expiration-phase image reconstruction using four bins and regularization factor values of 10-2.5 (1.50 mm) and 10-2.33 (1.35 mm) were found to optimize the tradeoff between sharpness, SNR, and artifacts. The optimized protocol facilitated clear visualization of the liver, blood vessels, and surrounding structures at isotropic resolutions of 1.50 and 1.35 mm in 3.5 min, without B1 + inhomogeneity effects in the shimmed liver region. A comparison between low-resolution fully sampled free-breathing (3.5 min) and breath-hold (19 s) acquisitions demonstrated comparable sharpness and SNR. To compare the 7 T data with 3 T MRI, 3 T scans were performed for two participants. 3 T reconstructions were done similarly to 7 T, excluding N4BiasFieldCorrection. Scan-specific regularization optimization was performed for fair comparison. Compared to 3 T, 7 T showed superior vascular contrast with inflow effects not observed at 3 T. Fold-over artifacts were present in 3 T scans but were minor at 7 T. 3 T and 7 T provided comparable results, with a much higher RF channel count at 3 T. In conclusion, high-resolution expiration-phase liver imaging at 7 T with homogeneous signal can be successfully achieved using a phase-shimmed, free-breathing protocol with a golden-angle pseudo-spiral sampling pattern technique and respiratory self-gating. This approach allows detailed anatomical depiction without the limitations of breath-holding, representing a significant advancement in ultra-high field abdominal MRI.

Introduction: Long term (>36 months) recovery after endovascular treatment (EVT) remains largely understudied in acute ischemic stroke (AIS) patients. Previously, we have shown that long-term white matter microstructure damage, iron deposition and atrophy within/remote to the 24-hour stroke lesion could influence long term quality of life. However, the role of baseline and 24-hour lesion characteristics on long term lesion characteristics and the influence of lesion status on recovery to perform daily activities is unknown. Methods: We executed a quantitative 3T MRI protocol (FLAIR, diffusion MRI, R1, R2*, Quantitative Susceptibility Mapping, Arterial Spin Labeling) after 37 (IQR: 28-38) months in patients from the MR CLEAN No-IV trial who received a 24-hour post-EVT MRI (DWI, FLAIR, SWI). Lateral ventricular volume difference was computed at both time-points to quantify atrophy. Long term MRI was also used to quantify iron deposition (R2*) and volume difference in subcortical nuclei relative to the contralateral nuclei. Long term recovery was defined using daily activity scores from EQ5D5 questionnaire. A Bayesian paired T-test of ventricular lateral atrophy between 24-hour and long term MRI was performed, and a Bayesian correlation of baseline clinical and 24-hour lesion characteristics with long term MRI outcomes (Bayes Factor (BF)<3: no/weak; BF>3: moderate; BF>20: strong evidence). Results: We included 24 patients with a 24-hour and long-term MRI protocol. Median age was 64 (51-75) years, 11 (46%) patients were randomized to receive intravenous thrombolysis (IVT) prior to EVT, and 19 (79%) patients had an MCA occlusion. Functional independence (mRS 0-2) at 90 days (21 (88%) patients) and long term (20 (83%) patients) was comparable. Ventricular lateral atrophy was observed (BF=4.10 5 ), and higher baseline blood pressure and poorer collaterals were associated with larger long term lateral ventricular atrophy (BF: 11, 11, 4.3 respectively). Lower ASPECTS and poorer collaterals were associated with higher thalamic volume difference (BF=111, 11 respectively). Lastly, lower hippocampal volume and iron deposition were associated with reduced long term daily activity (BF=26, BF=4). Conclusion: While functional independence in our cohort remains comparable over time, baseline, clinical, and treatment characteristics point to long-term brain degradation, which in turn impacts the ability of patients to perform daily activities.

BACKGROUND:Heightened reactivity in the amygdala measured by functional magnetic resonance imaging during emotional processing is considered a potential biomarker for clinical depression. Still, it is unknown whether this is also true for depressive symptoms in the general population, and - when in remission after recurrent depressive episodes - it is associated with future episodes. METHODS:Using the UK Biobank population study (n = 11,334), we investigated the association of amygdala reactivity during negative facial stimuli, focusing on lifetime depression (trait), depressive symptoms (state), and the modulating effect of antidepressant (AD) treatment thereof. We employed normative modeling (NM) to better incorporate population heterogeneity of the amygdala activity. RESULTS:In line with a previous study, depressive symptoms (state) over the last 2 weeks were not associated with the amygdala reactivity signal. Rather, our results indicate a significant positive association (p = 0.03, ω2 = 0.001) between amygdala response and the recurrence of depressive episodes (trait). Longitudinal analysis revealed that the group that had experienced a single depressive episode before showed a significantly increased amygdala response after additional episodes (p = 0.03, ω2 = 0.017). ADs were not associated with amygdala response directly, but decreased associations within episode recurrence severity. CONCLUSIONS:The amygdala response to negative stimuli was associated with an individual's risk of recurrence of depressive episodes, and AD treatment reduced these associations. This study highlights the relevance of amygdala reactivity as a trait, but not a state biomarker for (recurrent) depression. Moreover, it demonstrates the benefit of applying NM in the context of population data.

Many neurological diseases are characterized by the accumulation of toxic proteins in the brain. This accumulation has been associated with improper clearance from the parenchyma. Recent discoveries highlighted perivascular spaces, which are cerebrospinal fluid (CSF)-filled spaces, as the channels of brain clearance. The forces driving CSF mobility within perivascular spaces are still debated. Here we present a noninvasive, CSF-specific magnetic resonance imaging technique (CSF-Selective T2-prepared REadout with Acceleration and Mobility-encoding) that enables detailed in vivo measurement of CSF mobility in humans, down to the level of perivascular spaces located around penetrating vessels, which is close to protein production sites. We find region-specific drivers of CSF mobility and demonstrate that CSF mobility can be increased by entraining vasomotion. Furthermore, we find region-specific CSF mobility alterations in patients with cerebral amyloid angiopathy, a brain disorder associated with clearance impairment. The availability of this technique opens up avenues to investigate the impact of CSF-mediated clearance in neurodegeneration and sleep.

Background: Attention-Deficit/Hyperactivity Disorder (ADHD) is commonly treated with methylphenidate (MPH). Although highly effective, MPH treatment still has a relatively high non-response rate of around 30%, highlighting the need for a better understanding of treatment response. Radiomics of T1-weighted images and Diffusion Tensor Imaging (DTI) combined with machine learning approaches could offer a novel method for assessing MPH treatment response. Purpose: To evaluate the accuracy of both conventional and radiomics approaches in predicting treatment response based on baseline T1 and DTI data in stimulant-naive ADHD participants. Methods: We performed a secondary analysis of a randomized clinical trial (ePOD-MPH), involving 47 stimulant- naive ADHD participants (23 boys aged 11.4 +/- 0.4 years, 24 men aged 28.1 +/- 4.3 years) who underwent 16 weeks of treatment with MPH. Baseline T1-weighted and DTI MRI scans were acquired. Treatment response was assessed at 8 weeks (during treatment) and one week after cessation of 16-week treatment (post-treatment) using the Clinical Global Impressions- Improvement scale as our primary outcome. We compared prediction accuracy using a conventional model and a radiomics model. The conventional approach included the volume of bilateral caudate, putamen, pallidum, accumbens, and hippocampus, and for DTI the mean fractional anisotropy (FA) of the entire brain white matter, bilateral Anterior Thalamic Radiation (ATR), and the splenium of the corpus callosum, totaling 14 regional features. For the radiomics approach, 380 features (shape-based and first-order statistics) were extracted from these 14 regions. XGBoost models with nested cross-validation were used and constructed for the total cohort (n = 47), as well as children (n = 23) and adults (n = 24) separately. Exact binomial tests were employed to compare model performance. Results: For the conventional model, balanced accuracy (bAcc) in predicting treatment response during treatment was 63 % for the total cohort, 32 % for children, and 36 % for adults (Area Under the Receiver Operating Characteristic Curve (AUC-ROC): 0.69, 0.33, 0.41 respectively). Radiomics models demonstrated bAcc's of 68 %, 64 %, and 64 % during treatment (AUC-ROCs of 0.73, 0.62, 0.69 respectively). These predictions were better than chance for both conventional and radiomics models in the total cohort (p = 0.04, p = 0.003 respectively). The radiomics models outperformed the conventional models during treatment in children only (p = 0.02). At post-treatment, performance was markedly reduced. Conclusion: While conventional and radiomics models performed equally well in predicting clinical improvement across children and adults during treatment, radiomics features offered enhanced structural information beyond

INTRODUCTION:Major depressive disorder is a prevalent and debilitating disease. Finding effective treatment is a lengthy trial-and-error process. Thus, identifying predictors of treatment outcome is key to reducing disease burden. Brain turbulence-like dynamic measures on transmission of information might be a potential methodological strategy for identifying informative predictors of antidepressant treatment response. METHODS:We analyzed data of the EMBARC study in which 296 adult outpatients were randomized to eight weeks of sertraline or placebo treatment. Resting state functional MRI scans were acquired at baseline and one week after treatment initiation. Turbulence-like dynamic measures were computed. Generalized linear models were used to predict week 8 response and symptom severity scores. Primary performance metrics were the area under the receiver operating characteristic (AUROC) and the root mean squared error (RMSE) respectively. Permutation testing was performed to test for significance against chance. Internal leave-one-out cross-validation was performed on the sertraline-treated patients (sample A) and external validation on the placebo arm (sample B) and on placebo-treated, non-responding patients who later switched to sertraline (sample C). RESULTS:226 patients were analyzed (age 37.9±13.4 years; 148 [65.5%] female). Sample A included 109 participants, sample B 121, and sample C 61. Internal cross-validation results were significantly better than chance at predicting response (AUROC=0.71, balanced accuracy=69.5%, p=0.008) and symptom severity (RMSE=4.7, p=0.025). External validation in sample B did not yield performance significantly better than chance, and in sample C only prediction of response did (AUROC=0.65, balanced accuracy=60.3%, p=0.038). CONCLUSION:The turbulence framework is a suitable paradigm for predicting sertraline response in major depressive disorder. Prediction specificity to sertraline treatment was limited. Non-responders to sertraline were characterized by increased long-distance and reduced short-distance information cascade flow.

Accurately predicting individual antidepressant treatment response could expedite the lengthy trial-and-error process of finding an effective treatment for major depressive disorder (MDD). We tested and compared machine learning-based methods that predict individual-level pharmacotherapeutic treatment response using cortical morphometry from multisite longitudinal cohorts. We conducted an international analysis of pooled data from six sites of the ENIGMA-MDD consortium (n = 262 MDD patients; age = 36.5 ± 15.3 years; 154 (59%) female; mean response rate = 57%). Treatment response was defined as a ≥ 50% reduction in symptom severity score after 4-12 weeks post-initiation of antidepressant treatment. Structural MRI was acquired before, or < 14 days after, treatment initiation. The cortex was parcellated using FreeSurfer, from which cortical thickness and surface area were measured. We tested several machine learning pipeline configurations, which varied in (i) the way we presented the cortical data (i.e., average values per region of interest, as a vector containing voxel-wise cortical thickness and surface area measures, and as cortical thickness and surface area projections), (ii) whether we included clinical data, and the (iii) machine learning model (i.e., gradient boosting, support vector machine, and neural network classifiers) and (iv) cross-validation methods (i.e., k-fold and leave-one-site-out) we used. First, we tested if the overall predictive performance of the pipelines was better than chance, with a corrected 10-fold cross-validation permutation test. Second, we compared if some machine learning pipeline configurations outperformed others. In an exploratory analysis, we repeated our first analysis in three subpopulations, namely patients (i) from a single site, (ii) with comparable response rates, and (iii) showing the least (first quartile) and the most (fourth quartile) treatment response, which we call the extreme (non-)responders subpopulation. Finally, we explored the effect of including subcortical volumetric data on model performance. Overall, performance predicting antidepressant treatment response was not significantly better than chance (balanced accuracy = 50.5%; p = 0.66) and did not vary with alternative pipeline configurations. Exploratory analyses revealed that performance across models was only significantly better than chance in the extreme (non-)responders subpopulation (balanced accuracy = 63.9%, p = 0.001). Including subcortical data did not alter the observed model performance. Cortical structural MRI alone could not reliably predict individual pharmacotherapeutic treatment response in MDD. None of the used machine learning pipeline configurations outperformed the others. In exploratory analyses, we found that predicting response in the extreme (non-)responders subpopulation was feasible on both cortical data alone and combined with subcortical data, which suggests that specific MDD subpopulations may exhibit response-related patterns in structural data. Future work may use multimodal data to predict treatment response in MDD.

BackgroundWe retrospectively evaluated the quality of deep learning (DL) reconstructions of on-scanner accelerated intraoperative MRI (iMRI) during respective brain tumor surgery.MethodsAccelerated iMRI was performed using dual surface coils positioned around the area of resection. A DL model was trained on the fastMRI neuro dataset to mimic the data from the iMRI protocol. The evaluation was performed on imaging material from 40 patients imaged from Nov 1, 2021, to June 1, 2023, who underwent iMRI during tumor resection surgery. A comparative analysis was conducted between the conventional compressed sense (CS) method and the trained DL reconstruction method. Blinded evaluation of multiple image quality metrics was performed by two neuroradiologists and one neurosurgeon using a 1-to-5 Likert scale (1, nondiagnostic; 2, poor; 3, acceptable; 4, good; and 5, excellent), and the favored reconstruction variant.ResultsThe DL reconstruction was strongly favored or favored over the CS reconstruction for 33/40, 39/40, and 8/40 of cases for readers 1, 2, and 3, respectively. For the evaluation metrics, the DL reconstructions had a higher score than their respective CS counterparts for 72%, 72%, and 14% of the cases for readers 1, 2, and 3, respectively. Still, the DL reconstructions exhibited shortcomings such as a striping artifact and reduced signal.ConclusionDL shows promise in allowing for high-quality reconstructions of iMRI. The neuroradiologists noted an improvement in the perceived spatial resolution, signal-to-noise ratio, diagnostic confidence, diagnostic conspicuity, and spatial resolution compared to CS, while the neurosurgeon preferred the CS reconstructions across all metrics.Relevance statementDL shows promise to allow for high-quality reconstructions of iMRI, however, due to the challenging setting of iMRI, further optimization is needed.Key PointsiMRI is a surgical tool with a challenging image setting.DL allowed for high-quality reconstructions of iMRI.Additional optimization is needed due to the challenging intraoperative setting.

Background and Objectives: Artificial intelligence (AI) is revolutionizing Magnetic Resonance Imaging (MRI) along the acquisition and processing chain. Advanced AI frameworks have been applied in various successive tasks, such as image reconstruction, quantitative parameter map estimation, and image segmentation. However, existing frameworks are often designed to perform tasks independently of each other or are focused on specific models or single datasets, limiting generalization. This work introduces the Advanced Toolbox for Multitask Medical Imaging Consistency (ATOMMIC), a novel open-source toolbox that streamlines AI applications for accelerated MRI reconstruction and analysis. ATOMMIC implements several tasks using deep learning (DL) models and enables MultiTask Learning (MTL) to perform related tasks in an integrated manner, targeting generalization in the MRI domain. Methods: We conducted a comprehensive literature review and analyzed 12,479 GitHub repositories to assess the current landscape of AI frameworks for MRI. Subsequently, we demonstrate how ATOMMIC standardizes workflows and improves data interoperability, enabling effective benchmarking of various DL models across MRI tasks and datasets. To showcase ATOMMIC’s capabilities, we evaluated twenty-five DL models on eight publicly available datasets, focusing on accelerated MRI reconstruction, segmentation, quantitative parameter map estimation, and joint accelerated MRI reconstruction and segmentation using MTL. Results: ATOMMIC’s high-performance training and testing capabilities, utilizing multiple GPUs and mixed precision support, enable efficient benchmarking of multiple models across various tasks. The framework’s modular architecture implements each task through a collection of data loaders, models, loss functions, evaluation metrics, and pre-processing transformations, facilitating seamless integration of new tasks, datasets, and models. Our findings demonstrate that ATOMMIC supports MTL for multiple MRI tasks with harmonized complex-valued and real-valued data support while maintaining active development and documentation. Task-specific evaluations demonstrate that physics-based models outperform other approaches in reconstructing highly accelerated acquisitions. These high-quality reconstruction models also show superior accuracy in estimating quantitative parameter maps. Furthermore, when combining high-performing reconstruction models with robust segmentation networks through MTL, performance is improved in both tasks. Conclusions: ATOMMIC advances MRI reconstruction and analysis by leveraging MTL and ensuring consistency across tasks, models, and datasets. This comprehensive framework serves as a versatile platform for researchers to use existing AI methods and develop new approaches in medical imaging.

Purpose: To evaluate the quality of deep learning reconstruction for prospectively accelerated intraoperative magnetic resonance imaging (iMRI) during resective brain tumor surgery. Materials and Methods: Accelerated iMRI was performed during brain surgery using dual surface coils positioned around the area of resection. A deep learning (DL) model was trained on the fastMRI neuro dataset to mimic the data from the iMRI protocol. Evaluation was performed on imaging material from 40 patients imaged between 01.11.2021 - 01.06.2023 that underwent iMRI during tumor resection surgery. A comparative analysis was conducted between the conventional compressed sense (CS) method and the trained DL reconstruction method. Blinded evaluation of multiple image quality metrics was performed by two working neuro-radiologists and a working neurosurgeon on a 1 to 5 Likert scale (1=non diagnostic, 2=poor, 3=acceptable, 4=good, 5=excellent), and the favored reconstruction variant. Results: The DL reconstruction was strongly favored or favored over the CS reconstruction for 33/40, 39/40, and 8/40 of cases for reader 1, 2, and 3, respectively. Two of three readers consistently assigned higher ratings for the DL reconstructions, and the DL reconstructions had a higher score than their respective CS counterparts for 72%, 72%, and 14% of the cases for reader 1, 2, and 3, respectively. Still, the DL reconstructions exhibited shortcomings such as a striping artifact and reduced signal. Conclusion: DL shows promise to allow for high-quality reconstructions of intraoperative MRI with equal to or improved perceived spatial resolution, signal-to-noise ratio, diagnostic confidence, diagnostic conspicuity, and spatial resolution compared to compressed sense.

To obtain homogeneous signal throughout the human liver at 7 T. Flip angle (FA) shimming in 7T whole‐liver imaging was performed through parallel‐transmit kT‐point pulses based on subject‐specific multichannel absolute B1+$$ {\mathrm{B}}_1^{+} $$ maps from Fourier phase‐encoded dual refocusing echo acquisition mode (PE‐DREAM).

Morphological changes in the hippocampal, thalamic, and amygdala subfields have been suggested to form part of the pathophysiology of major depressive disorder (MDD). However, the use of conventional MRI scanners and acquisition techniques has prevented in-depth examinations at the subfield level, precluding a fine-grained understanding of these subfields and their involvement in MDD pathophysiology. We uniquely employed ultra-high field MRI at 7.0 Tesla to map hippocampal, thalamic, and amygdala subfields in MDD. Fifty-six MDD patients and 14 healthy controls (HCs) were enrolled in the final analysis. FreeSurfer protocols were used to segment hippocampal, thalamic, and amygdala subfields. Bayesian analysis was then implemented to assess differences between groups and relations with clinical features. While no effect was found for MDD diagnosis (i.e., case–control comparison), clinical characteristics of MDD patients were associated with subfield volumes of the hippocampus, thalamus, and amygdala. Specifically, the severity of depressive symptoms, insomnia, and childhood trauma in MDD patients related to lower thalamic subfield volumes. In addition, MDD patients with typical MDD versus those with atypical MDD showed lower hippocampal, thalamic, and amygdala subfield volumes. MDD patients with recurrent MDD versus those with first-episode MDD also showed lower thalamic subfield volumes. These findings allow uniquely fine-grained insights into hippocampal, thalamic, and amygdala subfield morphology in MDD, linking some of them to the clinical manifestation of MDD.

Whereas meta-analytical data highlight abnormal frontocortical macrostructure (thickness/surface area/volume) in Major Depressive Disorder (MDD), the underlying microstructural processes remain uncharted, due to the use of conventional MRI scanners and acquisition techniques. We uniquely combined Ultra-High Field MRI at 7.0 Tesla with Quantitative Imaging to map intracortical myelin (proxied by longitudinal relaxation time T1) and iron concentration (proxied by transverse relaxation time T2*), microstructural processes deemed particularly germane to cortical macrostructure. Informed by meta-analytical evidence, we focused specifically on orbitofrontal and rostral anterior cingulate cortices among adult MDD patients (N = 48) and matched healthy controls (HC; N = 10). Analyses probed the association of MDD diagnosis and clinical profile (severity, medication use, comorbid anxiety disorders, childhood trauma) with aforementioned microstructural properties. MDD diagnosis (p’s < 0.05, Cohen’s D = 0.55–0.66) and symptom severity (p’s < 0.01, r = 0.271–0.267) both related to decreased intracortical myelination (higher T1 values) within the lateral orbitofrontal cortex, a region tightly coupled to processing negative affect and feelings of sadness in MDD. No relations were found with local iron concentrations. These findings allow uniquely fine-grained insights on frontocortical microstructure in MDD, and cautiously point to intracortical demyelination as a possible driver of macroscale cortical disintegrity in MDD.

OBJECTIVE:Response to antidepressant treatment in major depressive disorder varies substantially between individuals, which lengthens the process of finding effective treatment. The authors sought to determine whether a multimodal machine learning approach could predict early sertraline response in patients with major depressive disorder. They assessed the predictive contribution of MR neuroimaging and clinical assessments at baseline and after 1 week of treatment.METHODS:This was a preregistered secondary analysis of data from the Establishing Moderators and Biosignatures of Antidepressant Response in Clinical Care (EMBARC) study, a multisite double-blind, placebo-controlled randomized clinical trial that included 296 adult outpatients with unmedicated recurrent or chronic major depressive disorder. MR neuroimaging and clinical data were collected before and after 1 week of treatment. Performance in predicting response and remission, collected after 8 weeks, was quantified using balanced accuracy (bAcc) and area under the receiver operating characteristic curve (AUROC) scores.RESULTS:A total of 229 patients were included in the analyses (mean age, 38 years [SD=13]; 66% female). Internal cross-validation performance in predicting response to sertraline (bAcc=68% [SD=10], AUROC=0.73 [SD=0.03]) was significantly better than chance. External cross-validation on data from placebo nonresponders (bAcc=62%, AUROC=0.66) and placebo nonresponders who were switched to sertraline (bAcc=65%, AUROC=0.68) resulted in differences that suggest specificity for sertraline treatment compared with placebo treatment. Finally, multimodal models outperformed unimodal models.CONCLUSIONS:The study results confirm that early sertraline treatment response can be predicted; that the models are sertraline specific compared with placebo; that prediction benefits from integrating multimodal MRI data with clinical data; and that perfusion imaging contributes most to these predictions. Using this approach, a lean and effective protocol could individualize sertraline treatment planning to improve psychiatric care.

To compare compressed sensing (CS) and the Cascades of Independently Recurrent Inference Machines (CIRIM) with respect to image quality and reconstruction times when 12-fold accelerated scans of patients with neurological deficits are reconstructed. Twelve-fold accelerated 3D T2-FLAIR images were obtained from a cohort of 62 patients with neurological deficits on 3 T MRI. Images were reconstructed offline via CS and the CIRIM. Image quality was assessed in a blinded and randomized manner by two experienced interventional neuroradiologists and one experienced pediatric neuroradiologist on imaging artifacts, perceived spatial resolution (sharpness), anatomic conspicuity, diagnostic confidence, and contrast. The methods were also compared in terms of self-referenced quality metrics, image resolution, patient groups and reconstruction time. In ten scans, the contrast ratio (CR) was determined between lesions and white matter. The effect of acceleration factor was assessed in a publicly available fully sampled dataset, since ground truth data are not available in prospectively accelerated clinical scans. Specifically, 451 FLAIR scans, including scans with white matter lesions, were adopted from the FastMRI database to evaluate structural similarity (SSIM) and the CR of lesions and white matter on ranging acceleration factors from four-fold up to 12-fold. Interventional neuroradiologists significantly preferred the CIRIM for imaging artifacts, anatomic conspicuity, and contrast. One rater significantly preferred the CIRIM in terms of sharpness and diagnostic confidence. The pediatric neuroradiologist preferred CS for imaging artifacts and sharpness. Compared to CS, the CIRIM reconstructions significantly improved in terms of imaging artifacts and anatomic conspicuity (p < 0.01) for higher resolution scans while yielding a 28

BACKGROUND:The thalamus serves as a central relay station within the brain, and thalamic connectional anomalies are increasingly thought to be present in major depressive disorder (MDD). However, the use of conventional MRI scanners and acquisition techniques has prevented a thorough examination of the thalamus and its subnuclear connectional profile. We combined ultra-high field diffusion MRI acquired at 7.0 Tesla to map the white matter connectivity of thalamic subnuclei. METHODS:Fifty-three MDD patients and 12 healthy controls (HCs) were involved in the final analysis. FreeSurfer was used to segment the thalamic subnuclei, and MRtrix was used to perform the preprocessing and tractography. Fractional anisotropy, axial diffusivity, mean diffusivity, radial diffusivity, and streamline count of thalamic subnuclear tracts were measured as proxies of white matter microstructure. Bayesian multilevel model was used to assess group differences in white matter metrics for each thalamic subnuclear tract and the association between these white matter metrics and clinical features in MDD. RESULTS:Evidence was found for reduced whiter matter metrics of the tracts spanning from all thalamic subnuclei among MDD versus HC participants. Moreover, evidence was found that white matter in various thalamic subnuclear tracts is related to medication status, age of onset and recurrence in MDD. CONCLUSIONS:Structural connectivity was generally reduced in thalamic subnuclei in MDD participants. Several clinical characteristics are related to perturbed subnuclear thalamic connectivity with cortical and subcortical circuits that govern sensory processing, emotional function, and goal-directed behavior.