Search

ObjectiveIn humans, arterial grayscale ultrasound texture features independently predict adverse cardiovascular disease (CVD) events and change with medical interventions. We performed this study to examine how grayscale ultrasound texture features and elastin fibers change in plaque-free segments of the arterial wall in a murine model prone to atherosclerosis.MethodsA total of 10 Apoetm1Unc/J mice (n = 5 male, n = 5 female) were imaged at 6, 16, and 24 weeks of age. Two mice were euthanized at 6 and 16 weeks and the remaining mice at 24 weeks. Texture features were extracted from the ultrasound images of the distal 1.0 mm of the common carotid artery wall, and elastin measures were extracted from histology images. Two-way analysis of variance was used to evaluate associations between week, sex, and grayscale texture features. Texture feature and elastin number comparisons between weeks were conducted using the sex-by-week two-way interaction contrasts. Sex-specific correlations between the number of elastin fibers and grayscale texture features were analyzed by conducting non-parametric Spearman's rank correlation analyses.ResultsArterial wall homogeneity changed significantly in male mice from 6 to 24 weeks, with a mean (SD) of 0.14 (0.03) units at 6 weeks and 0.18 (0.03) units at 24 weeks (p = 0.026). Spatial gray level dependence matrices-homogeneity (SGLD-HOM) also correlated with carotid artery plaque score (rs = 0.707, p = 0.033). Elastin fibers in the region of interest decreased from 6 to 24 weeks for both male and female mice, although only significantly in male mice. The mean (SD) number of elastin fibers for male mice was 5.32 (1.50) at 6 weeks and 3.59 (0.38) at 24 weeks (p = 0.023). For female mice, the mean (SD) number of elastin fibers was 3.98 (0.38) at 6 weeks and 3.46 (0.19) at 24 weeks (p = 0.051).ConclusionGrayscale ultrasound texture features that are associated with increased risk for CVD events in humans were used in a murine model, and the grayscale texture feature SGLD-HOM was shown to change in male mice from 6 weeks to 24 weeks. Structural alterations of the arterial wall (change in elastin fiber number) were observed during this time and may differ by sex.

There is significant interest in using neuroimaging data to predict behavior. The predictive models are often interpreted by the computation of feature importance, which quantifies the predictive relevance of an imaging feature. Tian and Zalesky (2021) suggest that feature importance estimates exhibit low split-half reliability, as well as a trade-off between prediction accuracy and feature importance reliability across parcellation resolutions. However, it is unclear whether the trade-off between prediction accuracy and feature importance reliability is universal. Here, we demonstrate that, with a sufficient sample size, feature importance (operationalized as Haufe-transformed weights) can achieve fair to excellent split-half reliability. With a sample size of 2600 participants, Haufe-transformed weights achieve average intra-class correlation coefficients of 0.75, 0.57 and 0.53 for cognitive, personality and mental health measures respectively. Haufe-transformed weights are much more reliable than original regression weights and univariate FC-behavior correlations. Original regression weights are not reliable even with 2600 participants. Intriguingly, feature importance reliability is strongly positively correlated with prediction accuracy across phenotypes. Within a particular behavioral domain, there is no clear relationship between prediction performance and feature importance reliability across regression models. Furthermore, we show mathematically that feature importance reliability is necessary, but not sufficient, for low feature importance error. In the case of linear models, lower feature importance error is mathematically related to lower prediction error. Therefore, higher feature importance reliability might yield lower feature importance error and higher prediction accuracy. Finally, we discuss how our theoretical results relate with the reliability of imaging features and behavioral measures. Overall, the current study provides empirical and theoretical insights into the relationship between prediction accuracy and feature importance reliability.

Sleep timing is controlled by intrinsic homeostatic and circadian components. The circadian component controls the chronotype, which is defined by the propensity to sleep at a particular clock time. However, sleep timing can be significantly affected by external factors such as the morning alarm clock. In this study, we analysed the timing of deep and REM sleep as well as the composition of REM sleep using Fitbit sleep staging in young healthy adults (n = 59) under real-life conditions. Sleep stage percentiles were correlated with the timing of total sleep in time after sleep onset for the homeostatic component and in clock time for the circadian component. Regarding the circadian component, the phase of total sleep is most strongly associated with the phases of early deep sleep and REM sleep. Furthermore, a stronger phase relationship between deep and REM sleep with total sleep is associated with greater consolidation of REM sleep. Chronotype-dependent sleep loss correlates negatively with the strength of the phase relationship between deep sleep and total sleep. In conclusion, the interaction of the circadian component of sleep timing with the timing of sleep stages is associated with REM sleep quality. In particular, the interaction of the circadian component of sleep timing with deep sleep seems to be more vulnerable to external factors.

The repertoire of currently available antiviral drugs spans therapeutic applications against a number of important human pathogens distributed worldwide. These include cases of the pandemic severe acute respiratory coronavirus type 2 (SARS-CoV-2 or COVID-19), human immunodeficiency virus type 1 (HIV-1 or AIDS), and the pregnancy- and posttransplant-relevant human cytomegalovirus (HCMV). In almost all cases, approved therapies are based on direct-acting antivirals (DAAs), but their benefit, particularly in long-term applications, is often limited by the induction of viral drug resistance or side effects. These issues might be addressed by the additional use of host-directed antivirals (HDAs). As a strong input from long-term experiences with cancer therapies, host protein kinases may serve as HDA targets of mechanistically new antiviral drugs. The study demonstrates such a novel antiviral strategy by targeting the major virus-supportive host kinase CDK7. Importantly, this strategy focuses on highly selective, 3D structure-derived CDK7 inhibitors carrying a warhead moiety that mediates covalent target binding. In summary, the main experimental findings of this study are as follows: (1) the in vitro verification of CDK7 inhibition and selectivity that confirms the warhead covalent-binding principle (by CDK-specific kinase assays), (2) the highly pronounced antiviral efficacies of the hit compounds (in cultured cell-based infection models) with half-maximal effective concentrations that reach down to picomolar levels, (3) a particularly strong potency of compounds against strains and reporter-expressing recombinants of HCMV (using infection assays in primary human fibroblasts), (4) additional activity against further herpesviruses such as animal CMVs and VZV, (5) unique mechanistic properties that include an immediate block of HCMV replication directed early (determined by Western blot detection of viral marker proteins), (6) a substantial drug synergism in combination with MBV (measured by a Loewe additivity fixed-dose assay), and (7) a strong sensitivity of clinically relevant HCMV mutants carrying MBV or ganciclovir resistance markers. Combined, the data highlight the huge developmental potential of this host-directed antiviral targeting concept utilizing covalently binding CDK7 inhibitors.

PurposeWe previously reported the safety and immunologic effects of a DNA vaccine (pTVG-HP [MVI-816]) encoding prostatic acid phosphatase (PAP) in patients with recurrent, nonmetastatic prostate cancer. The current trial evaluated the effects of this vaccine on metastatic progression.Patients and methodsNinety-nine patients with castration-sensitive prostate cancer and prostate-specific antigen (PSA) doubling time (DT) of less than 12 months were randomly assigned to treatment with either pTVG-HP co-administered intradermally with 200 μg granulocyte-macrophage colony-stimulating factor (GM-CSF) adjuvant or 200 μg GM-CSF alone six times at 14-day intervals and then quarterly for 2 years. The primary end point was 2-year metastasis-free survival (MFS). Secondary and exploratory end points were median MFS, changes in PSA DT, immunologic effects, and changes in quantitative 18F-sodium fluoride (NaF) positron emission tomography/computed tomography (PET/CT) imaging.ResultsTwo-year MFS was not different between study arms (41.8% vaccine v 42.3%; P = .97). Changes in PSA DT and median MFS were not different between study arms (18.9 v 18.3 months; hazard ratio [HR], 1.6; P = .13). Preplanned subset analysis identified longer MFS in vaccine-treated patients with rapid (< 3 months) pretreatment PSA DT (12.0 v 6.1 months; n = 21; HR, 4.4; P = .03). PAP-specific T cells were detected in both cohorts, including multifunctional PAP-specific T-helper 1-biased T cells. Changes in total activity (total standardized uptake value) on 18F-NaF PET/CT from months 3 to 6 increased 50% in patients treated with GM-CSF alone and decreased 23% in patients treated with pTVG-HP (n = 31; P = .07).ConclusionpTVG-HP treatment did not demonstrate an overall increase in 2-year MFS in patients with castration-sensitive prostate cancer, with the possible exception of a subgroup with rapidly progressive disease. Prespecified 18F-NaF PET/CT imaging conducted in a subset of patients suggests that vaccination had detectable effects on micrometastatic bone disease. Additional trials using pTVG-HP in combination with PD-1 blockade are under way.

BackgroundFearing increased myelotoxicity, many practitioners adjust the body surface area (BSA)-calculated doses in obese patients. Regarding temozolomide (TMZ), a prior study suggested men with a BSA >2 m2 may experience increased toxicity; however, surprisingly, the inverse observation was noted in women, ie, BSA

Background: Pulmonary hypertension causes pulmonary artery (PA) stiffening, which overloads the right ventricle (RV). Since symptoms of pulmonary hypertension (PH) are exacerbated by exercise, exercise-induced PA stiffening is relevant to cardiopulmonary status. Here, we sought to demonstrate the feasibility of using magnetic resonance imaging (MRI) for non-invasive assessment of exercise-induced changes in PA stiffness in patients with PH. Methods: MRI was performed on 7 PH patients and 8 age-matched control subjects at rest and during exercise stress. Main pulmonary artery (MPA) relative area change (RAC) and pulse wave velocity (PWV) were measured from 2D-PC images. Invasive right heart catheterization (RHC) was performed on 5 of the PH patients in conjunction with exercise stress to measure MPA pressures and stiffness index (β). Results: Heart rate and cardiac index (CI) were significantly increased with exercise in both groups. In controls, RAC decreased from 0.27 ± 0.05 at rest to 0.22 ± 0.06 with exercise (P < 0.05); a modest increase in PWV was not significant (P = 0.06). In PH patients, RAC decreased from 0.15 ± 0.02 to 0.11 ± 0.01 (P < 0.05) and PWV and β increased from 3.9 ± 0.54 m/s and 1.86 ± 0.12 at rest to 5.75 ± 0.70 m/s and 3.25 ± 0.26 with exercise (P < 0.05 for both), respectively. These results confirm increased MPA stiffness with exercise stress in both groups and the non-invasive metrics of MPA stiffness correlated well with β. Finally, as assessed by PWV but not RAC, PA stiffness of PH patients increased more than that of controls for comparable levels of moderate exercise. Conclusion: These results demonstrate the feasibility of using MRI for non-invasive assessment of exercise-induced changes in MPA stiffness in a small, heterogeneous group of PH patients in a research context. Similar measurements in a larger cohort are required to investigate differences between PWV and RAC for estimation of MPA stiffness.

Introduction: Despite its central role in medicine electrical stimulation (ES) is still limited by its selectivity. Different reports did assess effects of different waveforms, intensities, and frequency on the activation threshold of nerve fibres with different diameters. We aimed to extend this knowledge by investigating the effect of short monophasic rectangular pulses (1, 2, 5, 10, 50, 100 and 200 μs) on the recruitment order. Methods:The sciatic nerve of rats was stimulated, and the evoked compound nerve action potential (CNAP) measured at two sites on the tibialis nerve, using epineural electrodes. Changes in delay, amplitude, and the shape of the CNAP were analyzed. Results:The amplitude and delay of the CNAP were significantly affected by the pulse-width (PW). The delay and duration of the compound nerve action potential increased with longer PW, while the amplitude decreased. Discussion:Found changes are likely caused by changes in the time point of excitation of individual neuron fibres, depending on electrical field strength and exposure time. This might be of particular interest when selecting PWs for design and validation of stimulation patterns and analysis of experimental and clinical observations.

Abstract Background Pharmacist-led medication review and medication management programs (MMP) are well-known strategies to improve medication safety and effectiveness. If performed interprofessionally, outcomes might even improve. However, little is known about task sharing in interprofessional MMP, in which general practitioners (GPs) and community pharmacists (CPs) collaboratively perform medication reviews and continuously follow-up on patients with designated medical and pharmaceutical tasks, respectively. In 2016, ARMIN (Arzneimittelinitiative Sachsen-Thüringen) an interprofessional MMP was launched in two German federal states, Saxony and Thuringia. The aim of this study was to understand how GPs and CPs share tasks in MMP when reviewing the patients’ medication. Methods This was a cross-sectional postal survey among GPs and CPs who participated in the MMP. Participants were asked who completed which MMP tasks, e.g., checking drug-drug interactions, dosing, and side effects. In total, 15 MMP tasks were surveyed using a 5-point Likert scale ranging from “I complete this task alone” to “GP/CP completes this task alone”. The study was conducted between 11/2020 and 04/2021. Data was analyzed using descriptive statistics. Results In total, 114/165 (69.1%) GPs and 166/243 (68.3%) CPs returned a questionnaire. The majority of GPs and CPs reported (i) checking clinical parameters and medication overuse and underuse to be completed by GPs, (ii) checking storage conditions of drugs and initial compilation of the patient’s medication including brown bag review being mostly performed by CPs, and (iii) checking side-effects, non-adherence, and continuous updating of the medication list were carried out jointly. The responses differed most for problems with self-medication and adding and removing over-the-counter medicines from the medication list. In addition, the responses revealed that some MMP tasks were not sufficiently performed by either GPs or CPs. Conclusions Both GPs’ and CPs’ expertise are needed to perform MMP as comprehensively as possible. Future studies should explore how GPs and CPs can complement each other in MMP most efficiently.

Abstract The study of associations between inter-individual differences in brain structure and behaviour has a long history in psychology and neuroscience. Many associations between psychometric data, particularly intelligence and personality measures and local variations of brain structure have been reported. While the impact of such reported associations often goes beyond scientific communities, resonating in the public mind, their replicability is rarely evidenced. Previously, we have shown that associations between psychometric measures and estimates of grey matter volume (GMV) result in rarely replicated findings across large samples of healthy adults. However, the question remains if these observations are at least partly linked to the multidetermined nature of the variations in GMV, particularly within samples with wide age-range. Therefore, here we extended those evaluations and empirically investigated the replicability of associations of a broad range of psychometric variables and cortical thickness in a large cohort of healthy young adults. In line with our observations with GMV, our current analyses revealed low likelihood of significant associations and their rare replication across independent samples. We here discuss the implications of these findings within the context of accumulating evidence of the general poor replicability of structural-brain-behaviour associations, and more broadly of the replication crisis.

Ice nucleation proteins have the same tandemly arrayed water-organizing motifs seen in some antifreeze proteins, but on a larger scale. The authors show that mutation, interruption, and truncation of these arrays reduce ice nucleation activity indicating that the two protein types share a common mechanism.

The human frontal operculum (FOp) is a brain region that covers parts of the ventral frontal cortex next to the insula. Functional imaging studies showed activations in this region in tasks related to language, somatosensory, and cognitive functions. While the precise cytoarchitectonic areas that correlate to these processes have not yet been revealed, earlier receptorarchitectonic analysis resulted in a detailed parcellation of the FOp. We complemented this analysis by a cytoarchitectonic study of a sample of ten postmortem brains and mapped the posterior FOp in serial, cell-body stained histological sections using image analysis and multivariate statistics. Three new areas were identified: Op5 represents the most posterior area, followed by Op6 and the most anterior region Op7. Areas Op5-Op7 approach the insula, up to the circular sulcus. Area 44 of Broca’s region, the most ventral part of premotor area 6, and parts of the parietal operculum are dorso-laterally adjacent to Op5-Op7. The areas did not show any interhemispheric or sex differences. Three-dimensional probability maps and a maximum probability map were generated in stereotaxic space, and then used, in a first proof-of-concept-study, for functional decoding and analysis of structural and functional connectivity. Functional decoding revealed different profiles of cytoarchitectonically identified Op5-Op7. While left Op6 was active in music cognition, right Op5 was involved in chewing/swallowing and sexual processing. Both areas showed activation during the exercise of isometric force in muscles. An involvement in the coordination of flexion/extension could be shown for the right Op6. Meta-analytic connectivity modeling revealed various functional connections of the FOp areas within motor and somatosensory networks, with the most evident connection with the music/language network for Op6 left. The new cytoarchitectonic maps are part of Julich-Brain, and publicly available to serve as a basis for future analyses of structural-functional relationships in this region.

ObjectiveDelirium is associated with worse outcomes in patients with stroke and neurocritical illness, but delirium detection in these patients can be challenging with existing screening tools. To address this gap, we aimed to develop and evaluate machine learning models that detect episodes of post-stroke delirium based on data from wearable activity monitors in conjunction with stroke-related clinical features.DesignProspective observational cohort study.SettingNeurocritical Care and Stroke Units at an academic medical center.PatientsWe recruited 39 patients with moderate-to-severe acute intracerebral hemorrhage (ICH) and hemiparesis over a 1-year period [mean (SD) age 71.3 (12.20), 54% male, median (IQR) initial NIH Stroke Scale 14.5 (6), median (IQR) ICH score 2 (1)].Measurements and main resultsEach patient received daily assessments for delirium by an attending neurologist, while activity data were recorded throughout each patient's hospitalization using wrist-worn actigraph devices (on both paretic and non-paretic arms). We compared the predictive accuracy of Random Forest, SVM and XGBoost machine learning methods in classifying daily delirium status using clinical information alone and combined with actigraph data. Among our study cohort, 85% of patients (n = 33) had at least one delirium episode, while 71% of monitoring days (n = 209) were rated as days with delirium. Clinical information alone had a low accuracy in detecting delirium on a day-to-day basis [accuracy mean (SD) 62% (18%), F1 score mean (SD) 50% (17%)]. Prediction performance improved significantly (p < 0.001) with the addition of actigraph data [accuracy mean (SD) 74% (10%), F1 score 65% (10%)]. Among actigraphy features, night-time actigraph data were especially relevant for classification accuracy.ConclusionsWe found that actigraphy in conjunction with machine learning models improves clinical detection of delirium in patients with stroke, thus paving the way to make actigraph-assisted predictions clinically actionable.

Resting-state fMRI is commonly used to derive brain parcellations, which are widely used for dimensionality reduction and interpreting human neuroscience studies. We previously developed a model that integrates local and global approaches for estimating areal-level cortical parcellations. The resulting local-global parcellations are often referred to as the Schaefer parcellations. However, the lack of homotopic correspondence between left and right Schaefer parcels has limited their use for brain lateralization studies. Here, we extend our previous model to derive homotopic areal-level parcellations. Using resting-fMRI and task-fMRI across diverse scanners, acquisition protocols, preprocessing and demographics, we show that the resulting homotopic parcellations are as homogeneous as the Schaefer parcellations, while being more homogeneous than five publicly available parcellations. Furthermore, weaker correlations between homotopic parcels are associated with greater lateralization in resting network organization, as well as lateralization in language and motor task activation. Finally, the homotopic parcellations agree with the boundaries of a number of cortical areas estimated from histology and visuotopic fMRI, while capturing sub-areal (e.g., somatotopic and visuotopic) features. Overall, these results suggest that the homotopic local-global parcellations represent neurobiologically meaningful subdivisions of the human cerebral cortex and will be a useful resource for future studies. Multi-resolution parcellations estimated from 1479 participants are publicly available (https://github.com/ThomasYeoLab/CBIG/tree/master/stable_projects/brain_parcellation/Yan2023_homotopic).

Resting-state functional connectivity (RSFC) is widely used to predict behavioral measures. To predict behavioral measures, representing RSFC with parcellations and gradients are the two most popular approaches. Here, we compare parcellation and gradient approaches for RSFC-based prediction of a broad range of behavioral measures in the Human Connectome Project (HCP) and Adolescent Brain Cognitive Development (ABCD) datasets. Among the parcellation approaches, we consider group-average “hard” parcellations (Schaefer et al., 2018), individual-specific “hard” parcellations (Kong et al., 2021a), and an individual-specific “soft” parcellation (spatial independent component analysis with dual regression; Beckmann et al., 2009). For gradient approaches, we consider the well-known principal gradients (Margulies et al., 2016) and the local gradient approach that detects local RSFC changes (Laumann et al., 2015). Across two regression algorithms, individual-specific hard-parcellation performs the best in the HCP dataset, while the principal gradients, spatial independent component analysis and group-average “hard” parcellations exhibit similar performance. On the other hand, principal gradients and all parcellation approaches perform similarly in the ABCD dataset. Across both datasets, local gradients perform the worst. Finally, we find that the principal gradient approach requires at least 40 to 60 gradients to perform as well as parcellation approaches. While most principal gradient studies utilize a single gradient, our results suggest that incorporating higher order gradients can provide significant behaviorally relevant information. Future work will consider the inclusion of additional parcellation and gradient approaches for comparison.

Naturalistic viewing (NV) is currently considered a promising paradigm for studying individual differences in functional brain organization. While whole brain functional connectivity (FC) under NV has been relatively well characterized, so far little work has been done on a network level.Here, we extend current knowledge by characterizing the influence of NV on FC in fourteen meta-analytically derived brain networks considering three different movie stimuli in comparison to resting-state (RS). We show that NV increases identifiability of individuals over RS based on functional connectivity in certain, but not all networks. Furthermore, movie stimuli including a narrative appear more distinct from RS. In addition, we assess individual variability in network FC by comparing within- and between-subject similarity during NV and RS. We show that NV can evoke individually distinct NFC patterns by increasing inter-subject variability while retaining within-subject similarity. Crucially, our results highlight that this effect is not observable across all networks, but rather dependent on the network-stimulus combination. Our results confirm that NV can improve the detection of individual differences over RS and underline the importance of selecting the appropriate combination of movie and cognitive network for the research question at hand.

As part of the RiPoD study, a large cohort (n = 700) of new mothers were recruited 1-6 days postpartum at the university hospital in Aachen and were observed for a period of 12 weeks. While 9% of them were found to develop postpartum depression (PPD), 13% developed postpartum adjustment disorder (AD). The PPD and AD cases were accurately identified by means of a combination of clinical and remote assessment tools (Hahn et al. 2021). In the exploration cohort, a balanced accuracy of 87% was achieved in week 3 postpartum using a combination of baseline and follow-up depression EPDS scores and mood scores (online mood assessment) for the differentiation between PPD and ND. The results were confirmed in the independent replication cohort. Within the first 6 weeks postpartum, the clinical manifestation of AD and PPD resembled each other based on personal and family history of depression, stressful life events, depressive symptoms, and attachment to the child. A comparison between cumulative hair cortisol and cortisone levels in the third trimester of pregnancy and the first 12 weeks postpartum revealed that only ND and AD (and not PPD) had pregnancy-related physiological adaptation in cortisol/cortisone levels, which reverted to the pre-pregnancy baseline following delivery (Stickel et al. 2021). Approximately 2 days after delivery, we also found that women who developed either AD or PPD within 12 weeks did not differ from ND immediately after delivery on the basis of structural and functional neuroimaging data (Schnakenberg et al. 2021). In the very early postpartum period, the mothers showed a significant reduction in gray matter volume in brain areas related to socio-cognitive and emotional processes compared to their nulliparous counterparts, indicating an effect of pregnancy on brain structure (Chechko et al. 2021).

The difference between age predicted using anatomical brain scans and chronological age, i.e., the brain-age delta, provides a proxy for atypical aging. Various data representations and machine learning (ML) algorithms have been used for brain-age estimation. However, how these choices compare on performance criteria important for real-world applications, such as; (1) within-dataset accuracy, (2) cross-dataset generalization, (3) test-retest reliability, and (4) longitudinal consistency, remains uncharacterized. We evaluated 128 workflows consisting of 16 feature representations derived from gray matter (GM) images and eight ML algorithms with diverse inductive biases. Using four large neuroimaging databases covering the adult lifespan (total N = 2953, 18–88 years), we followed a systematic model selection procedure by sequentially applying stringent criteria. The 128 workflows showed a within-dataset mean absolute error (MAE) between 4.73–8.38 years, from which 32 broadly sampled workflows showed a cross-dataset MAE between 5.23–8.98 years. The test-retest reliability and longitudinal consistency of the top 10 workflows were comparable. The choice of feature representation and the ML algorithm both affected the performance. Specifically, voxel-wise feature spaces (smoothed and resampled), with and without principal components analysis, with non-linear and kernel-based ML algorithms performed well. Strikingly, the correlation of brain-age delta with behavioral measures disagreed between within-dataset and cross-dataset predictions. Application of the best-performing workflow on the ADNI sample showed a significantly higher brain-age delta in Alzheimer's and mild cognitive impairment patients compared to healthy controls. However, in the presence of age bias, the delta estimates in the patients varied depending on the sample used for bias correction. Taken together, brain-age shows promise, but further evaluation and improvements are needed for its real-world application.

Deep neural networks currently provide the most advanced and accurate machine learning models to distinguish between structural MRI scans of subjects with Alzheimer’s disease and healthy controls. Unfortunately, the subtle brain alterations captured by these models are difficult to interpret because of the complexity of these multi-layer and non-linear models. Several heatmap methods have been proposed to address this issue and analyze the imaging patterns extracted from the deep neural networks, but no quantitative comparison between these methods has been carried out so far. In this work, we explore these questions by deriving heatmaps from Convolutional Neural Networks (CNN) trained using T1 MRI scans of the ADNI data set and by comparing these heatmaps with brain maps corresponding to Support Vector Machine (SVM) activation patterns. Three prominent heatmap methods are studied: Layer-wise Relevance Propagation (LRP), Integrated Gradients (IG), and Guided Grad-CAM (GGC). Contrary to prior studies where the quality of heatmaps was visually or qualitatively assessed, we obtained precise quantitative measures by computing overlap with a ground-truth map from a large meta-analysis that combined 77 voxel-based morphometry (VBM) studies independently from ADNI. Our results indicate that all three heatmap methods were able to capture brain regions covering the meta-analysis map and achieved better results than SVM activation patterns. Among them, IG produced the heatmaps with the best overlap with the independent meta-analysis.