Harmonization of multi-site diffusion tensor imaging data

Resolving heterogeneity in Alzheimer's disease based on individualized structural covariance network

The heterogeneity of Alzheimer's disease (AD) poses a challenge to precision medicine. We aimed to identify distinct subtypes of AD based on the individualized structural covariance network (IDSCN) analysis and to research the underlying neurobiology mechanisms. In this study, 187 patients with AD (age = 73.57 ± 6.00, 50% female) and 143 matched normal controls (age = 74.30 ± 7.80, 44% female) were recruited from the Alzheimer's Disease Neuroimaging Initiative (ADNI) project database, and T1 images were acquired. We utilized the IDSCN analysis to generate individual-level altered structural covariance network and performed k-means clustering to subtype AD based on structural covariance network. Cognition, disease progression, morphological features, and gene expression profiles were further compared between subtypes, to characterize the heterogeneity in AD. Two distinct AD subtypes were identified in a reproducible manner, and we named the two subtypes as slow progression type (subtype 1, n = 104, age = 76.15 ± 6.44, 42% female) and rapid progression type (subtype 2, n = 83, age = 71.98 ± 8.72, 47% female), separately. Subtype 1 had better baseline visuospatial function than subtype 2 (p < 0.05), whereas subtype 2 had better baseline memory function than subtype 1 (p < 0.05). Subtype 2 showed worse progression in memory (p = 0.003), language (p = 0.003), visuospatial function (p = 0.020), and mental state (p = 0.038) than subtype 1. Subtype 1 often shared increased structural covariance network, mainly in the frontal lobe and temporal lobe regions, whereas subtype 2 often shared increased structural covariance network, mainly in occipital lobe regions and temporal lobe regions. Functional annotation further revealed that all differential structural covariance network between the two AD subtypes were mainly implicated in memory, learning, emotion, and cognition. Additionally, differences in gray matter volume (GMV) between AD subtypes were identified, and genes associated with GMV differences were found to be enriched in the terms potassium ion transport, synapse organization, and histone modification and the pathways viral infection, neurodegeneration-multiple diseases, and long-term depression. The two distinct AD subtypes were identified and characterized with neuroanatomy, cognitive trajectories, and gene expression profiles. These comprehensive results have implications for neurobiology mechanisms and precision medicine.

Contrastive image adaptation for acquisition shift reduction in medical imaging

The domain shift, or acquisition shift in medical imaging, is responsible for potentially harmful differences between development and deployment conditions of medical image analysis techniques. There is a growing need in the community for advanced methods that could mitigate this issue better than conventional approaches. In this paper, we consider configurations in which we can expose a learning-based pixel level adaptor to a large variability of unlabeled images during its training, i.e. sufficient to span the acquisition shift expected during the training or testing of a downstream task model. We leverage the ability of convolutional architectures to efficiently learn domain-agnostic features and train a many-to-one unsupervised mapping between a source collection of heterogeneous images from multiple unknown domains subjected to the acquisition shift and a homogeneous subset of this source set of lower cardinality, potentially constituted of a single image. To this end, we propose a new cycle-free image-to-image architecture based on a combination of three loss functions : a contrastive PatchNCE loss, an adversarial loss and an edge preserving loss allowing for rich domain adaptation to the target image even under strong domain imbalance and low data regimes. Experiments support the interest of the proposed contrastive image adaptation approach for the regularization of downstream deep supervised segmentation and cross-modality synthesis models.

Thalamic nuclei volume differences in schizophrenia patients and healthy controls using probabilistic mapping: A comparative analysis

AIM

We aimed to investigate potential discrepancies in the volume of thalamic nuclei between individuals with schizophrenia and healthy controls.

METHODS

The imaging data for this study were obtained from the MCICShare data repository within SchizConnect. We employed probabilistic mapping technique developed by Iglesias et al. (2018). The analytical component entailed volumetric segmentation of the thalamus using the FreeSurfer image analysis suite. Our analysis focused on evaluating the differences in the volumes of various thalamic nuclei groups within the thalami, specifically the anterior, intralaminar, medial, posterior, lateral, and ventral groups in both the right and left thalami, between schizophrenia patients and healthy controls. We employed MANCOVA to analyse these dependent variables (volumes of 12 distinct thalamic nuclei groups), with diagnosis (SCZ vs. HCs) as the main explanatory variable, while controlling for covariates such as eTIV and age.

RESULTS

The assumptions of MANCOVA, including the homogeneity of covariance matrices, were met. Specific univariate tests for the right thalamus revealed significant differences in the medial (F[1, 200] = 26.360, p < 0.001), and the ventral groups (F[1, 200] = 4.793, p = 0.030). For the left thalamus, the medial (F[1, 200] = 22.527, p < 0.001); posterior (F[1, 200] = 8.227, p = 0.005), lateral (F[1, 200] = 7.004, p = 0.009), and ventral groups (F[1, 200] = 9.309, p = 0.003) showed significant differences.

CONCLUSION

These findings suggest that particular thalamic nuclei groups in both the right and left thalami may be most affected in schizophrenia, with more pronounced differences observed in the left thalamic nuclei.

FUNDINGS

The authors received no financial support for the research.

Examining the association between posttraumatic stress disorder and disruptions in cortical networks identified using data-driven methods

Posttraumatic stress disorder (PTSD) is associated with lower cortical thickness (CT) in prefrontal, cingulate, and insular cortices in diverse trauma-affected samples. However, some studies have failed to detect differences between PTSD patients and healthy controls or reported that PTSD is associated with greater CT. Using data-driven dimensionality reduction, we sought to conduct a well-powered study to identify vulnerable networks without regard to neuroanatomic boundaries. Moreover, this approach enabled us to avoid the excessive burden of multiple comparison correction that plagues vertex-wise methods. We derived structural covariance networks (SCNs) by applying non-negative matrix factorization (NMF) to CT data from 961 PTSD patients and 1124 trauma-exposed controls without PTSD. We used regression analyses to investigate associations between CT within SCNs and PTSD diagnosis (with and without accounting for the potential confounding effect of trauma type) and symptom severity in the full sample. We performed additional regression analyses in subsets of the data to examine associations between SCNs and comorbid depression, childhood trauma severity, and alcohol abuse. NMF identified 20 unbiased SCNs, which aligned closely with functionally defined brain networks. PTSD diagnosis was most strongly associated with diminished CT in SCNs that encompassed the bilateral superior frontal cortex, motor cortex, insular cortex, orbitofrontal cortex, medial occipital cortex, anterior cingulate cortex, and posterior cingulate cortex. CT in these networks was significantly negatively correlated with PTSD symptom severity. Collectively, these findings suggest that PTSD diagnosis is associated with widespread reductions in CT, particularly within prefrontal regulatory regions and broader emotion and sensory processing cortical regions.

A Physics-Based Algorithm to Universally Standardize Routinely Obtained Clinical T2-Weighted Images

RATIONALE AND OBJECTIVES

MR images can be challenging for machine learning and other large-scale analyses because most clinical images, for example, T2-weighted (T2w) images, reflect not only the biologically relevant T2 of tissue but also hardware and acquisition parameters that vary from site to site. Quantitative T2 mapping avoids these confounds because it quantitatively isolates the biological parameter of interest, thus representing a universal standardization across sites. However, efforts to incorporate quantitative mapping sequences into routine clinical practice have seen slow adoption. Here we show, for the first time, that the routine T2w complex raw dataset can be successfully regarded as a quantitative mapping sequence that can be reconstructed with classical optimization methods and physics-based constraints.

MATERIALS AND METHODS

While previous constrained reconstruction methods are unable to reconstruct a T2 map based on this data, the expanding-constrained alternating minimization for parameter mapping (e-CAMP), which employs stepwise initialization, a linearized version of the exponential model and a phase conjugacy constraint, is demonstrated to provide useful quantitative maps directly from a vendor T2w single image data.

RESULTS

This paper introduces the method and demonstrates its performance using simulations, retrospectively undersampled brain images, and prospectively acquired T2w images taken on both phantom and brain.

CONCLUSION

Because T2w scans are included in nearly every protocol, this approach could open the door to creating large, standardized datasets without requiring widespread changes in clinical protocols.

Neurite-based white matter alterations in MAPT mutation carriers: A multi-shell diffusion MRI study in the ALLFTD consortium

We assessed white matter (WM) integrity in MAPT mutation carriers (16 asymptomatic, 5 symptomatic) compared to 31 non-carrier family controls using diffusion tensor imaging (DTI) (fractional anisotropy; FA, mean diffusivity; MD) and neurite orientation dispersion and density imaging (NODDI) (neurite density index; NDI, orientation and dispersion index; ODI). Linear mixed-effects models accounting for age and family relatedness revealed alterations across DTI and NODDI metrics in all mutation carriers and in symptomatic carriers, with the most significant differences involving fronto-temporal WM tracts. Asymptomatic carriers showed higher entorhinal MD and lower cingulum FA and patterns of higher ODI mostly involving temporal areas and long association and projections fibers. Regression models between estimated time to or time from disease and DTI and NODDI metrics in key regions (amygdala, cingulum, entorhinal, inferior temporal, uncinate fasciculus) in all carriers showed increasing abnormalities with estimated time to or time from disease onset, with FA and NDI showing the strongest relationships. Neurite-based metrics, particularly ODI, appear to be particularly sensitive to early WM involvement in asymptomatic carriers.

Impaired segregation of the attention deficit hyperactivity disorder related pattern in children

BACKGROUND

Inattention is a key characteristic of attention deficit hyperactivity disorder (ADHD). Specific brain abnormalities associated with this symptom form a discernible pattern related with ADHD in children (i.e., ADHD related pattern) in our earlier research. The developmental processes of segregation and integration may be crucial to ADHD. However, how brains reconfigure these processes of the ADHD related pattern in different subtypes of ADHD and across sexes remain unclear.

METHODS

Nested-spectral partition method was applied to identify effects of subtype and sex on segregation and integration of the ADHD related pattern, using 145 ADHD patients and 135 typically developing controls (TDC) aged 7-14. Relationships between the measures and inattention symptoms were also investigated.

RESULTS

Children with ADHD exhibited lower segregation of the ADHD related pattern (p = 1.17 × 10-8) than TDCs. Only the main effect of subtype was significant (p = 1.14 × 10-5). Both ADHD-C (p = 2.16 × 10-6) and ADHD-I (p = 2.87 × 10-6) patients had lower segregation components relative to the TDC. Moreover, segregation components were negatively correlated with inattention scores.

CONCLUSIONS

This study identified impaired segregation in the ADHD related pattern of children with ADHD and found shared neural bases among different subtypes and sexes.

Cortical microstructural associations with CSF amyloid and pTau

Diffusion MRI (dMRI) can be used to probe microstructural properties of brain tissue and holds great promise as a means to non-invasively map Alzheimer's disease (AD) pathology. Few studies have evaluated multi-shell dMRI models such as neurite orientation dispersion and density imaging (NODDI) and mean apparent propagator (MAP)-MRI in cortical gray matter where many of the earliest histopathological changes occur in AD. Here, we investigated the relationship between CSF pTau181 and Aβ1-42 burden and regional cortical NODDI and MAP-MRI indices in 46 cognitively unimpaired individuals, 18 with mild cognitive impairment, and two with dementia (mean age: 71.8 ± 6.2 years) from the Alzheimer's Disease Neuroimaging Initiative. We compared findings to more conventional cortical thickness measures. Lower CSF Aβ1-42 and higher pTau181 were associated with cortical dMRI measures reflecting less hindered or restricted diffusion and greater diffusivity. Cortical dMRI measures, but not cortical thickness measures, were more widely associated with Aβ1-42 than pTau181 and better distinguished Aβ+ from Aβ- participants than pTau+ from pTau- participants. dMRI associations mediated the relationship between CSF markers and delayed logical memory performance, commonly impaired in early AD. dMRI metrics sensitive to early AD pathogenesis and microstructural damage may be better measures of subtle neurodegeneration in comparison to standard cortical thickness and help to elucidate mechanisms underlying cognitive decline.

Identifying psychosis subtypes use individualized covariance structural differential networks and multi-site clustering

BACKGROUND

Similarities among schizophrenia (SZ), schizoaffective disorder (SAD) and bipolar disorder (BP) including clinical phenotypes, brain alterations and risk genes, make it challenging to perform reliable separation among them. However, previous subtype identification that transcend traditional diagnostic boundaries were based on group-level neuroimaging features, ignoring individual-level inferences.

METHODS

455 psychoses (178 SZs, 134 SADs and 143 BPs), their first-degree relatives (N = 453) and healthy controls (HCs, N = 220) were collected from Bipolar-Schizophrenia Network on Intermediate Phenotypes (B-SNIP I) consortium. Individualized covariance structural differential networks (ICSDNs) were constructed for each patient and multi-site clustering was used to identify psychosis subtypes. Group differences between subtypes in clinical phenotypes and voxel-wise fractional amplitude of low frequency fluctuation (fALFF) were calculated, as well as between the corresponding relatives.

RESULTS

Two psychosis subtypes were identified with increased whole brain structural covariance, with decreased connectivity between amygdala-hippocampus and temporal-occipital cortex in subtype I (S-I) compared to subtype II (S-II), which was replicated under different clustering methods, number of edges and across datasets (B-SNIP II) and different brain atlases. S-I had higher emotional-related symptoms than S-II and showed significant fALFF decrease in temporal and occipital cortex, while S-II was more similar to HC. This pattern was consistently validated on relatives of S-I and S-II in both fALFF and clinical symptoms.

CONCLUSIONS

These findings reconcile categorical and dimensional perspectives of psychosis neurobiological heterogeneity, indicating that relatives of S-I might have greater predisposition in developing psychosis, while relatives of S-II are more likely to be healthy. This study contributes to the development of neuroimaging informed diagnostic classifications within psychosis spectrum.

White matter and literacy: A dynamic system in flux

Cross-sectional studies have linked differences in white matter tissue properties to reading skills. However, past studies have reported a range of, sometimes conflicting, results. Some studies suggest that white matter properties act as individual-level traits predictive of reading skill, whereas others suggest that reading skill and white matter develop as a function of an individual's educational experience. In the present study, we tested two hypotheses: a) that diffusion properties of the white matter reflect stable brain characteristics that relate to stable individual differences in reading ability or b) that white matter is a dynamic system, linked with learning over time. To answer these questions, we examined the relationship between white matter and reading in a five-year longitudinal dataset and a series of large-scale, single-observation, cross-sectional datasets (N = 14,249 total participants). We find that gains in reading skill correspond to longitudinal changes in the white matter. However, in the cross-sectional datasets, we find no evidence for the hypothesis that individual differences in white matter predict reading skill. These findings highlight the link between dynamic processes in the white matter and learning.

Improving brain age prediction with anatomical feature attention-enhanced 3D-CNN

Currently, significant progress has been made in predicting brain age from structural Magnetic Resonance Imaging (sMRI) data using deep learning techniques. However, despite the valuable structural information they contain, the traditional engineering features known as anatomical features have been largely overlooked in this context. To address this issue, we propose an attention-based network design that integrates anatomical and deep convolutional features, leveraging an anatomical feature attention (AFA) module to effectively capture salient anatomical features. In addition, we introduce a fully convolutional network, which simplifies the extraction of deep convolutional features and overcomes the high computational memory requirements associated with deep learning. Our approach outperforms several widely-used models on eight publicly available datasets (n = 2501), with a mean absolute error (MAE) of 2.20 years in predicting brain age. Comparisons with deep learning models lacking the AFA module demonstrate that our fusion model effectively improves overall performance. These findings provide a promising approach for combining anatomical and deep convolutional features from sMRI data to predict brain age, with potential applications in clinical diagnosis and treatment, particularly for populations with age-related cognitive decline or neurological disorders.

The magnitude and direction of the relationship between risk factor and cognition depends on age: a pooled analysis of 5 community-based studies

The mixed evidence of the association between high levels of cardiovascular risk factors (CVRF) and the risk for cognitive impairment may be due to confounding of age across studies. We pooled and harmonized individual-level data (30,967 persons, age range 42-96 years) from five prospective cohorts to investigate by 1 year age increments to investigate whether or not there is change in slope describing the association of CVRF to a cognitive outcome (Digit Symbol Substitution Test; DSST). The CVRF included: systolic and diastolic blood pressure, total cholesterol, fasting glucose and body mass index. Linear and quadratic piecewise regression models were fit to the trajectory patterns of these slopes (betas). The pattern of yearly slope changes showed higher CVRF were associated with lower DSST, but associations attenuated toward zero as age increased for all but DBP where 1 year slopes for DBP changed direction from negative to positive from mid- to late-age. Age is not only a driver of cognitive decline-age also modifies the direction and strength of the association of cognitive function to CVRF and cohort age may be one reason why the evidence for CVRF-CD association is mixed.

Untargeted metabolomic profiling in children identifies novel pathways in asthma and atopy

BACKGROUND

Asthma and other atopic disorders can present with varying clinical phenotypes marked by differential metabolomic manifestations and enriched biological pathways.

OBJECTIVE

We sought to identify these unique metabolomic profiles in atopy and asthma.

METHODS

We analyzed baseline nonfasted plasma samples from a large multisite pediatric population of 470 children aged <13 years from 3 different sites in the United States and France. Atopy positivity (At+) was defined as skin prick test result of ≥3 mm and/or specific IgE ≥ 0.35 IU/mL and/or total IgE ≥ 173 IU/mL. Asthma positivity (As+) was based on physician diagnosis. The cohort was divided into 4 groups of varying combinations of asthma and atopy, and 6 pairwise analyses were conducted to best assess the differential metabolomic profiles between groups.

RESULTS

Two hundred ten children were classified as At-As-, 42 as At+As-, 74 as At-As+, and 144 as At+As+. Untargeted global metabolomic profiles were generated through ultra-high-performance liquid chromatography-tandem mass spectroscopy. We applied 2 independent machine learning classifiers and short-listed 362 metabolites as discriminant features. Our analysis showed the most diverse metabolomic profile in the At+As+/At-As- comparison, followed by the At-As+/At-As- comparison, indicating that asthma is the most discriminant condition associated with metabolomic changes. At+As+ metabolomic profiles were characterized by higher levels of bile acids, sphingolipids, and phospholipids, and lower levels of polyamine, tryptophan, and gamma-glutamyl amino acids.

CONCLUSION

The At+As+ phenotype displays a distinct metabolomic profile suggesting underlying mechanisms such as modulation of host-pathogen and gut microbiota interactions, epigenetic changes in T-cell differentiation, and lower antioxidant properties of the airway epithelium.

Learning site-invariant features of connectomes to harmonize complex network measures

Multi-site diffusion MRI data is often acquired on different scanners and with distinct protocols. Differences in hardware and acquisition result in data that contains site dependent information, which confounds connectome analyses aiming to combine such multi-site data. We propose a data-driven solution that isolates site-invariant information whilst maintaining relevant features of the connectome. We construct a latent space that is uncorrelated with the imaging site and highly correlated with patient age and a connectome summary measure. Here, we focus on network modularity. The proposed model is a conditional, variational autoencoder with three additional prediction tasks: one for patient age, and two for modularity trained exclusively on data from each site. This model enables us to 1) isolate site-invariant biological features, 2) learn site context, and 3) re-inject site context and project biological features to desired site domains. We tested these hypotheses by projecting 77 connectomes from two studies and protocols (Vanderbilt Memory and Aging Project (VMAP) and Biomarkers of Cognitive Decline Among Normal Individuals (BIOCARD) to a common site. We find that the resulting dataset of modularity has statistically similar means (p-value <0.05) across sites. In addition, we fit a linear model to the joint dataset and find that positive correlations between age and modularity were preserved.

Evaluation of Mean Shift, ComBat, and CycleGAN for Harmonizing Brain Connectivity Matrices Across Sites

Connectivity matrices derived from diffusion MRI (dMRI) provide an interpretable and generalizable way of understanding the human brain connectome. However, dMRI suffers from inter-site and between-scanner variation, which impedes analysis across datasets to improve robustness and reproducibility of results. To evaluate different harmonization approaches on connectivity matrices, we compared graph measures derived from these matrices before and after applying three harmonization techniques: mean shift, ComBat, and CycleGAN. The sample comprises 168 age-matched, sex-matched normal subjects from two studies: the Vanderbilt Memory and Aging Project (VMAP) and the Biomarkers of Cognitive Decline Among Normal Individuals (BIOCARD). First, we plotted the graph measures and used coefficient of variation (CoV) and the Mann-Whitney U test to evaluate different methods' effectiveness in removing site effects on the matrices and the derived graph measures. ComBat effectively eliminated site effects for global efficiency and modularity and outperformed the other two methods. However, all methods exhibited poor performance when harmonizing average betweenness centrality. Second, we tested whether our harmonization methods preserved correlations between age and graph measures. All methods except for CycleGAN in one direction improved correlations between age and global efficiency and between age and modularity from insignificant to significant with p-values less than 0.05.

Merging or ensembling: integrative analysis in multiple neuroimaging studies

The aim of this paper is to systematically investigate merging and ensembling methods for spatially varying coefficient mixed effects models (SVCMEM) in order to carry out integrative learning of neuroimaging data obtained from multiple biomedical studies. The "merged" approach involves training a single learning model using a comprehensive dataset that encompasses information from all the studies. Conversely, the "ensemble" approach involves creating a weighted average of distinct learning models, each developed from an individual study. We systematically investigate the prediction accuracy of the merged and ensemble learners under the presence of different degrees of interstudy heterogeneity. Additionally, we establish asymptotic guidelines for making strategic decisions about when to employ either of these models in different scenarios, along with deriving optimal weights for the ensemble learner. To validate our theoretical results, we perform extensive simulation studies. The proposed methodology is also applied to 3 large-scale neuroimaging studies.

Evaluation of Mean Shift, ComBat, and CycleGAN for Harmonizing Brain Connectivity Matrices Across Sites

Connectivity matrices derived from diffusion MRI (dMRI) provide an interpretable and generalizable way of understanding the human brain connectome. However, dMRI suffers from inter-site and between-scanner variation, which impedes analysis across datasets to improve robustness and reproducibility of results. To evaluate different harmonization approaches on connectivity matrices, we compared graph measures derived from these matrices before and after applying three harmonization techniques: mean shift, ComBat, and CycleGAN. The sample comprises 168 age-matched, sex-matched normal subjects from two studies: the Vanderbilt Memory and Aging Project (VMAP) and the Biomarkers of Cognitive Decline Among Normal Individuals (BIOCARD). First, we plotted the graph measures and used coefficient of variation (CoV) and the Mann-Whitney U test to evaluate different methods' effectiveness in removing site effects on the matrices and the derived graph measures. ComBat effectively eliminated site effects for global efficiency and modularity and outperformed the other two methods. However, all methods exhibited poor performance when harmonizing average betweenness centrality. Second, we tested whether our harmonization methods preserved correlations between age and graph measures. All methods except for CycleGAN in one direction improved correlations between age and global efficiency and between age and modularity from insignificant to significant with p-values less than 0.05.

Efficacy of MRI data harmonization in the age of machine learning: a multicenter study across 36 datasets

Pooling publicly-available MRI data from multiple sites allows to assemble extensive groups of subjects, increase statistical power, and promote data reuse with machine learning techniques. The harmonization of multicenter data is necessary to reduce the confounding effect associated with non-biological sources of variability in the data. However, when applied to the entire dataset before machine learning, the harmonization leads to data leakage, because information outside the training set may affect model building, and potentially falsely overestimate performance. We propose a 1) measurement of the efficacy of data harmonization; 2) harmonizer transformer, i.e., an implementation of the ComBat harmonization allowing its encapsulation among the preprocessing steps of a machine learning pipeline, avoiding data leakage by design. We tested these tools using brain T1-weighted MRI data from 1740 healthy subjects acquired at 36 sites. After harmonization, the site effect was removed or reduced, and we showed the data leakage effect in predicting individual age from MRI data, highlighting that introducing the harmonizer transformer into a machine learning pipeline allows for avoiding data leakage by design.

White matter integrity in bipolar disorder investigated with diffusion tensor magnetic resonance imaging and fractal geometry

BACKGROUND

Growing evidence suggests the presence of white matter (WM) alterations in bipolar disorder (BD). In this study we aimed to investigate the state of WM structures, in terms of tissue integrity and morphological complexity, in BD patients compared to healthy controls (HC), in an attempt to better elucidate the microstructural changes associated with BD.

METHODS

We collected a dataset of 399 Diffusion Tensor Magnetic Resonance Imaging (167 BD and 232 healthy controls) images, acquired at five different sites, which was processed with Tract-Based Spatial Statistics (TBSS) and fractal analysis.

RESULTS

The TBSS analysis demonstrated significantly lower FA values in the BD group. Diffusion abnormalities were primarily located in the temporo-parietal network. The Fractal Dimension (FD) analysis did not reveal consistent significant differences in the morphological complexity of WM structures between the groups. When the FD values of patients were considered individually, it is possible to notice some localized significant deviations from the healthy population.

LIMITATIONS

DTI sequences have not been harmonized before acquisition, samples' sizes are heterogeneous.

CONCLUSIONS

This study, by applying both TBSS and FD analyses, allows to evaluate diffusion and structural alterations of WM at the same time. The evaluation of WM integrity from these two different perspectives could be useful to better understand the pathophysiological and morphological changes underpinning bipolar disorder.

Multi-site benchmark classification of major depressive disorder using machine learning on cortical and subcortical measures

Machine learning (ML) techniques have gained popularity in the neuroimaging field due to their potential for classifying neuropsychiatric disorders. However, the diagnostic predictive power of the existing algorithms has been limited by small sample sizes, lack of representativeness, data leakage, and/or overfitting. Here, we overcome these limitations with the largest multi-site sample size to date (N = 5365) to provide a generalizable ML classification benchmark of major depressive disorder (MDD) using shallow linear and non-linear models. Leveraging brain measures from standardized ENIGMA analysis pipelines in FreeSurfer, we were able to classify MDD versus healthy controls (HC) with a balanced accuracy of around 62%. But after harmonizing the data, e.g., using ComBat, the balanced accuracy dropped to approximately 52%. Accuracy results close to random chance levels were also observed in stratified groups according to age of onset, antidepressant use, number of episodes and sex. Future studies incorporating higher dimensional brain imaging/phenotype features, and/or using more advanced machine and deep learning methods may yield more encouraging prospects.

Deep learning based joint fusion approach to exploit anatomical and functional brain information in autism spectrum disorders

BACKGROUND

The integration of the information encoded in multiparametric MRI images can enhance the performance of machine-learning classifiers. In this study, we investigate whether the combination of structural and functional MRI might improve the performances of a deep learning (DL) model trained to discriminate subjects with Autism Spectrum Disorders (ASD) with respect to typically developing controls (TD).

MATERIAL AND METHODS

We analyzed both structural and functional MRI brain scans publicly available within the ABIDE I and II data collections. We considered 1383 male subjects with age between 5 and 40 years, including 680 subjects with ASD and 703 TD from 35 different acquisition sites. We extracted morphometric and functional brain features from MRI scans with the Freesurfer and the CPAC analysis packages, respectively. Then, due to the multisite nature of the dataset, we implemented a data harmonization protocol. The ASD vs. TD classification was carried out with a multiple-input DL model, consisting in a neural network which generates a fixed-length feature representation of the data of each modality (FR-NN), and a Dense Neural Network for classification (C-NN). Specifically, we implemented a joint fusion approach to multiple source data integration. The main advantage of the latter is that the loss is propagated back to the FR-NN during the training, thus creating informative feature representations for each data modality. Then, a C-NN, with a number of layers and neurons per layer to be optimized during the model training, performs the ASD-TD discrimination. The performance was evaluated by computing the Area under the Receiver Operating Characteristic curve within a nested 10-fold cross-validation. The brain features that drive the DL classification were identified by the SHAP explainability framework.

RESULTS

The AUC values of 0.66±0.05 and of 0.76±0.04 were obtained in the ASD vs. TD discrimination when only structural or functional features are considered, respectively. The joint fusion approach led to an AUC of 0.78±0.04. The set of structural and functional connectivity features identified as the most important for the two-class discrimination supports the idea that brain changes tend to occur in individuals with ASD in regions belonging to the Default Mode Network and to the Social Brain.

CONCLUSIONS

Our results demonstrate that the multimodal joint fusion approach outperforms the classification results obtained with data acquired by a single MRI modality as it efficiently exploits the complementarity of structural and functional brain information.

Learning multi-site harmonization of magnetic resonance images without traveling human phantoms

Harmonization improves Magn. Reson. Imaging (MRI) data consistency and is central to effective integration of diverse imaging data acquired across multiple sites. Recent deep learning techniques for harmonization are predominantly supervised in nature and hence require imaging data of the same human subjects to be acquired at multiple sites. Data collection as such requires the human subjects to travel across sites and is hence challenging, costly, and impractical, more so when sufficient sample size is needed for reliable network training. Here we show how harmonization can be achieved with a deep neural network that does not rely on traveling human phantom data. Our method disentangles site-specific appearance information and site-invariant anatomical information from images acquired at multiple sites and then employs the disentangled information to generate the image of each subject for any target site. We demonstrate with more than 6,000 multi-site T1- and T2-weighted images that our method is remarkably effective in generating images with realistic site-specific appearances without altering anatomical details. Our method allows retrospective harmonization of data in a wide range of existing modern large-scale imaging studies, conducted via different scanners and protocols, without additional data collection.

Microstructural Organization of Distributed White Matter Associated With Fine Motor Control in US Service Members With Mild Traumatic Brain Injury

Mild traumatic brain injury (mTBI) is the most common form of brain injury. While most individuals recover from mTBI, roughly 20% experience persistent symptoms, potentially including reduced fine motor control. We investigate relationships between regional white matter organization and subcortical volumes associated with performance on the Grooved Pegboard (GPB) test in a large cohort of military Service Members and Veterans (SM&Vs) with and without a history of mTBI(s). Participants were enrolled in the Long-term Impact of Military-relevant Brain Injury Consortium-Chronic Effects of Neurotrauma Consortium. SM&Vs with a history of mTBI(s) (n = 847) and without mTBI (n = 190) underwent magnetic resonance imaging and the GPB test. We first examined between-group differences in GPB completion time. We then investigated associations between GPB performance and regional structural imaging measures (tractwise diffusivity, subcortical volumes, and cortical thickness) in SM&Vs with a history of mTBI(s). Lastly, we explored whether mTBI history moderated associations between imaging measures and GPB performance. SM&Vs with mTBI(s) performed worse than those without mTBI(s) on the non-dominant hand GPB test at a trend level (p < 0.1). Higher fractional anisotropy (FA) of tracts including the posterior corona radiata, superior longitudinal fasciculus, and uncinate fasciculus were associated with better GPB performance in the dominant hand in SM&Vs with mTBI(s). These findings support that the organization of several white matter bundles are associated with fine motor performance in SM&Vs. We did not observe that mTBI history moderated associations between regional FA and GPB test completion time, suggesting that chronic mTBI may not significantly influence fine motor control.

Riemannian frameworks for the harmonization of resting-state functional MRI scans

Magnetic Resonance Imaging provides unprecedented images of the brain. Unfortunately, scanners and acquisition protocols can significantly impact MRI scans. The development of statistical methods able to reduce this variability without altering the relevant information in the scans, often coined harmonization methods, has been the topic of an increasing research effort supported by the recent growth of publicly available neuroimaging data sets and new possibilities for combining them to achieve greater statistical power. In this work, we focus on the challenges specifically raised by the harmonization of resting-state functional MRI scans. We propose to harmonize resting-state fMRI scans by reducing the impact of covariates such as scanner differences and scanning protocols on their associated functional connectomes and then propagating the changes back to the rs-fMRI time series. We use Riemannian geometric frameworks to preserve the mathematical properties of functional connectomes during their harmonization, and we demonstrate how state-of-the-art harmonization methods can be embedded within these frameworks to reduce covariates effects while preserving the relevant clinical information associated with aging or brain disorders. During our experiments, a large set of synthetic data was generated and processed to compare eighty variants of the proposed approach. The framework achieving the best harmonization was then applied to three low-dimensional data sets made of 712 sets of fMRI time series provided by the ABIDE consortium and two high-dimensional data sets obtained by processing 1527 rs-fMRI scans provided by the Human Connectome Project, the Framingham Heart Study and the Genetics of Brain Structure and Function study. These experiments established that our new framework could successfully harmonize low-dimensional connectomes and voxelwise functional time series and confirmed the need for preserving connectomes properties during their harmonization.

Association of white matter microstructural alteration with non-suicidal self-injury behavior and visual working memory in adolescents with borderline personality disorder

BACKGROUND

Non-suicidal self-injurious behavior (NSSI) is the core characteristic of adolescent borderline personality disorder (BPD) and visual working memory is involved in the pathological processes of BPD. This study aimed to investigate alterations in white matter microstructure and their association with NSSI and visual working memory in adolescents with BPD.

METHODS

53 adolescents diagnosed with BPD and 39 healthy controls (HCs) were enrolled. White matter microstructure was assessed with the fractional anisotropy (FA) and mean diffusivity (MD) from diffusion tensor imaging (DTI). Correlation analysis was performed to assess the association between FA/MD and core features of BPD. A mediation analysis was performed to test whether the effects of white matter alterations on NSSI could be mediated by visual working memory.

RESULTS

Adolescents with BPD showed a reduced FA and an increased MD in the cortical-limbic and cortical-thalamus circuit when compared to the HCs (p < 0.05). Increased MD was positively correlated with NSSI, impulse control and identity disturbance (p < 0.05), and was negatively correlated with the score of visual reproduction. Reserved visual working memory masked the effects of white matter microstructural alterations on NSSI behavior.

CONCLUSIONS

White matter microstructural deficits in the cortical-limbic and cortical-thalamus circuits may be associated with NSSI and visual working memory in adolescents with BPD. Reserved visual working memory may protect against NSSI.

State of the art: radiomics and radiomics-related artificial intelligence on the road to clinical translation

Radiomics and artificial intelligence carry the promise of increased precision in oncologic imaging assessments due to the ability of harnessing thousands of occult digital imaging features embedded in conventional medical imaging data. While powerful, these technologies suffer from a number of sources of variability that currently impede clinical translation. In order to overcome this impediment, there is a need to control for these sources of variability through harmonization of imaging data acquisition across institutions, construction of standardized imaging protocols that maximize the acquisition of these features, harmonization of post-processing techniques, and big data resources to properly power studies for hypothesis testing. For this to be accomplished, it will be critical to have multidisciplinary and multi-institutional collaboration.

Current Auditory Hallucinations Are Not Associated With Specific White Matter Diffusion Alterations in Schizophrenia

Background and Hypothesis

Studies have linked auditory hallucinations (AH) in schizophrenia spectrum disorders (SCZ) to altered cerebral white matter microstructure within the language and auditory processing circuitry (LAPC). However, the specificity to the LAPC remains unclear. Here, we investigated the relationship between AH and DTI among patients with SCZ using diffusion tensor imaging (DTI).

Study Design

We included patients with SCZ with (AH+; n = 59) and without (AH-; n = 81) current AH, and 140 age- and sex-matched controls. Fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD) were extracted from 39 fiber tracts. We used principal component analysis (PCA) to identify general factors of variation across fiber tracts and DTI metrics. Regression models adjusted for sex, age, and age2 were used to compare tract-wise DTI metrics and PCA factors between AH+, AH-, and healthy controls and to assess associations with clinical characteristics.

Study Results

Widespread differences relative to controls were observed for MD and RD in patients without current AH. Only limited differences in 2 fiber tracts were observed between AH+ and controls. Unimodal PCA factors based on MD, RD, and AD, as well as multimodal PCA factors, differed significantly relative to controls for AH-, but not AH+. We did not find any significant associations between PCA factors and clinical characteristics.

Conclusions

Contrary to previous studies, DTI metrics differed mainly in patients without current AH compared to controls, indicating a widespread neuroanatomical distribution. This challenges the notion that altered DTI metrics within the LAPC is a specific feature underlying AH.

Robust fiber orientation distribution function estimation using deep constrained spherical deconvolution for diffusion-weighted magnetic resonance imaging

Purpose

Diffusion-weighted magnetic resonance imaging (DW-MRI) is a critical imaging method for capturing and modeling tissue microarchitecture at a millimeter scale. A common practice to model the measured DW-MRI signal is via fiber orientation distribution function (fODF). This function is the essential first step for the downstream tractography and connectivity analyses. With recent advantages in data sharing, large-scale multisite DW-MRI datasets are being made available for multisite studies. However, measurement variabilities (e.g., inter- and intrasite variability, hardware performance, and sequence design) are inevitable during the acquisition of DW-MRI. Most existing model-based methods [e.g., constrained spherical deconvolution (CSD)] and learning-based methods (e.g., deep learning) do not explicitly consider such variabilities in fODF modeling, which consequently leads to inferior performance on multisite and/or longitudinal diffusion studies.

Approach

In this paper, we propose a data-driven deep CSD method to explicitly constrain the scan-rescan variabilities for a more reproducible and robust estimation of brain microstructure from repeated DW-MRI scans. Specifically, the proposed method introduces a three-dimensional volumetric scanner-invariant regularization scheme during the fODF estimation. We study the Human Connectome Project (HCP) young adults test-retest group as well as the MASiVar dataset (with inter- and intrasite scan/rescan data). The Baltimore Longitudinal Study of Aging dataset is employed for external validation.

Results

From the experimental results, the proposed data-driven framework outperforms the existing benchmarks in repeated fODF estimation. By introducing the contrastive loss with scan/rescan data, the proposed method achieved a higher consistency while maintaining higher angular correlation coefficients with the CSD modeling. The proposed method is assessing the downstream connectivity analysis and shows increased performance in distinguishing subjects with different biomarkers.

Conclusion

We propose a deep CSD method to explicitly reduce the scan-rescan variabilities, so as to model a more reproducible and robust brain microstructure from repeated DW-MRI scans. The plug-and-play design of the proposed approach is potentially applicable to a wider range of data harmonization problems in neuroimaging.

Ultrasound Radiomics for the Prediction of Breast Cancers with HER2-Zero, -Low, and -Positive Status: A Dual-Center Study

PURPOSE

To assess whether gray-scale ultrasound (US) based radiomic features can help distinguish HER2 expressions (ie, HER2-overexpressing, HER2-low-expressing, and HER2-zero-expressing) in breast cancer.

MATERIALS AND METHODS

This retrospective study encompassed female breast cancer patients who underwent US examinations at two distinct centers from February 2021 to July 2023. Tumor segmentation and radiomic feature extraction were performed on grayscale US images. Decision Tree analysis was employed to simultaneously evaluate feature importance, and the Least Absolute Shrinkage and Selection Operator technique was utilized for feature selection to construct the radiomic signature. The Area Under the Curve (AUC) of the Receiver Operating Characteristic curve was employed to assess the performance of the radiomic features. Multivariate logistic regression was used to identify independent predictors for distinguishing HER2 expression in the dataset.

RESULTS

The training set comprised 292 patients from Center 1 (median, 51 years; interquartile range [IQR]: 45-61), while the external validation set included 131 patients from Center 2 (median, 51 years; IQR: 45-62). In the external validation dataset, the radiomic features achieved AUC of 0.76 for distinguishing between HER2-low and positive tumors versus HER2-zero tumors. The AUC for differentiating HER2-low (1+) from HER2-zero tumors was 0.74, and for distinguishing HER2-low (2+) from HER2-zero tumors, the AUC was 0.77. In the multivariate analysis assessing HER2-low and HER2-positive versus HER2-zero tumors, internal echoes (P = .029) and margins (P < .001) emerged as independent predictive factors.

CONCLUSION

The radiomic signature and tumor descriptors from gray-scale US may predict distinct HER2 expressions of breast cancers with therapeutic implications.

Functional connectome hierarchy dysfunction in Alzheimer's disease and its relationship with cognition and gene expression profiling

Numerous researches have shown that the human brain organizes as a continuum axis crossing from sensory motor to transmodal cortex. Functional network alterations were commonly found in Alzheimer's disease (AD). Whether the hierarchy of AD brain networks has changed and how these changes related to gene expression profiling and cognition is unclear. Using resting-state functional magnetic resonance imaging data from 233 subjects (185 AD patients and 48 healthy controls), we studied the changes in the functional network gradients in AD. Moreover, we investigated the relationships between gradient alterations and cognition, and gene expression profiling, respectively. We found that the second gradient organizes as a continuum axis crossing from the sensory motor to the transmodal cortex. Compared to the healthy controls, the secondary gradient scores of the visual and somatomotor network (SOM) increased significantly in AD, and the secondary gradient scores of default mode and frontoparietal network decreased significantly in AD. The secondary gradient scores of SOM and salience network (SAL) significantly positively correlated with memory function in AD. The secondary gradient in SAL also significantly positively correlated with language function. The AD-related second gradient alterations were spatially associated with the gene expression and the relevant genes enriched in neurobiology-related pathways, specially expressed in various tissues, cell types, and developmental stages. These findings suggested the changes in the functional network gradients in AD and deepened our understanding of the correlation between macroscopic gradient structure and microscopic gene expression profiling in AD.

Internalizing Symptoms and Adverse Childhood Experiences Associated With Functional Connectivity in a Middle Childhood Sample

BACKGROUND

Research has found overlapping associations in adults of resting-state functional connectivity (RSFC) to both internalizing disorders (e.g., depression, anxiety) and a history of traumatic events. The present study aimed to extend this previous research to a younger sample by examining RSFC associations with both internalizing symptoms and adverse childhood experiences (ACEs) in middle childhood.

METHODS

We used generalized linear mixed models to examine associations between a priori within- and between-network RSFC with child-reported internalizing symptoms and ACEs using the Adolescent Brain Cognitive Development dataset (N = 10,168, mean age = 9.95 years, SD = 0.627).

RESULTS

We found that internalizing symptoms and ACEs were associated with both multiple overlapping and unique RSFC network patterns. Both ACEs and internalizing symptoms were associated with a reduced anticorrelation between the default mode network and the dorsal attention network. However, internalizing symptoms were uniquely associated with lower within-network default mode network connectivity, while ACEs were uniquely associated with both lower between-network connectivity of the auditory network and cingulo-opercular network, and higher within-network frontoparietal network connectivity.

CONCLUSIONS

The present study points to overlap in the RSFC associations with internalizing symptoms and ACEs, as well as important areas of specificity in RSFC associations. Many of the RSFC associations found have been previously implicated in attentional control functions, including modulation of attention to sensory stimuli. This may have critical importance in understanding internalizing symptoms and outcomes of ACEs.

Scanner-Independent MyoMapNet for Accelerated Cardiac MRI T1 Mapping Across Vendors and Field Strengths

BACKGROUND

In cardiac T1 mapping, a series of T1 -weighted (T1 w) images are collected and numerically fitted to a two or three-parameter model of the signal recovery to estimate voxel-wise T1 values. To reduce the scan time, one can collect fewer T1 w images, albeit at the cost of precision or/and accuracy. Recently, the feasibility of using a neural network instead of conventional two- or three-parameter fit modeling has been demonstrated. However, prior studies used data from a single vendor and field strength; therefore, the generalizability of the models has not been established.

PURPOSE

To develop and evaluate an accelerated cardiac T1 mapping approach based on MyoMapNet, a convolution neural network T1 estimator that can be used across different vendors and field strengths by incorporating the relevant scanner information as additional inputs to the model.

STUDY TYPE

Retrospective, multicenter.

POPULATION

A total of 1423 patients with known or suspected cardiac disease (808 male, 57 ± 16 years), from three centers, two vendors (Siemens, Philips), and two field strengths (1.5 T, 3 T). The data were randomly split into 60% training, 20% validation, and 20% testing.

FIELD STRENGTH/SEQUENCE

A 1.5 T and 3 T, Modified Look-Locker inversion recovery (MOLLI) for native and postcontrast T1 .

ASSESSMENT

Scanner-independent MyoMapNet (SI-MyoMapNet) was developed by altering the deep learning (DL) architecture of MyoMapNet to incorporate scanner vendor and field strength as inputs. Epicardial and endocardial contours and blood pool (by manually drawing a large region of interest in the blood pool) of the left ventricle were manually delineated by three readers, with 2, 8, and 9 years of experience, and SI-MyoMapNet myocardial and blood pool T1 values (calculated from four T1 w images) were compared with conventional MOLLI T1 values (calculated from 8 to 11 T1 w images).

STATISTICAL TESTS

Equivalency test with 95% confidence interval (CI), linear regression slope, Pearson correlation coefficient (r), Bland-Altman analysis.

RESULTS

The proposed SI-MyoMapNet successfully created T1 maps. Native and postcontrast T1 values measured from SI-MyoMapNet were strongly correlated with MOLLI, despite using only four T1 w images, at both field-strengths and vendors (all r > 0.86). For native T1 , SI-MyoMapNet and MOLLI were in good agreement for myocardial and blood T1 values in institution 1 (myocardium: 5 msec, 95% CI [3, 8]; blood: -10 msec, 95%CI [-16, -4]), in institution 2 (myocardium: 6 msec, 95% CI [0, 11]; blood: 0 msec, [-18, 17]), and in institution 3 (myocardium: 7 msec, 95% CI [-8, 22]; blood: 8 msec, [-14, 30]). Similar results were observed for postcontrast T1 .

DATA CONCLUSION

Inclusion of field strength and vendor as additional inputs to the DL architecture allows generalizability of MyoMapNet across different vendors or field strength.

EVIDENCE LEVEL

2.

TECHNICAL EFFICACY

Stage 2.

Prediction of H3K27M alteration in midline gliomas of the brain using radiomics: A multi-institute study

Background

Noninvasive prediction of H3K27M-altered Diffuse midline gliomas is important because of the involvement of deep locations and proximity to eloquent structures. We aim to predict H3K27M alteration in midline gliomas using radiomics features of T2W images.

Methods

Radiomics features extracted from 124 subjects (69 H3K27M-altered/55 H3K27M-wild type). T2W images were resampled to 1 × 1 × 1mm3 voxel size, preprocessed, and normalized for artifact correction, intensity variations. The feature set was normalized and subjected to reduction by variance thresholding, correlation coefficient thresholding, and sequential feature selector. Adaptive synthesis oversampling technique was used to oversample the training data. Random forest classifier (RFC), Decision tree classifier (DTC), and K-nearest neighbors classifier (KNN) were trained over the training dataset and the performance was assessed over the internal test dataset and external test data set (52 subjects: 33 H3K27M-altered/19-H3K27M-wild type).

Results

DTC achieved a validation score of 77.33% (5-fold cross-validation) and an accuracy of 80.64%, 75% on internal and external test datasets. RFC achieved a validation score of 80.7% (5-fold cross-validation) an accuracy of 80.6%, and 73% on internal and external test datasets. DTC achieved a validation score of 78.67% (5-fold cross-validation) an accuracy of 80.64%, and 61.53% on internal and external test datasets. The accuracy score of DTC, RFC, and KNN on the internal test dataset was approximately 80% while on the external test dataset, DTC achieved 75% accuracy, RFC achieved 73% accuracy and KNN achieved 65.1% accuracy.

Conclusions

H3K27M alteration is a potential immunotherapeutic marker and is associated with poor prognosis and radiomics features extracted from conventional T2W-images can help in identifying H3K27M-altered cases non-invasively with high precision.

The effects of data leakage on connectome-based machine learning models

Predictive modeling has now become a central technique in neuroimaging to identify complex brain-behavior relationships and test their generalizability to unseen data. However, data leakage, which unintentionally breaches the separation between data used to train and test the model, undermines the validity of predictive models. Previous literature suggests that leakage is generally pervasive in machine learning, but few studies have empirically evaluated the effects of leakage in neuroimaging data. Although leakage is always an incorrect practice, understanding the effects of leakage on neuroimaging predictive models provides insight into the extent to which leakage may affect the literature. Here, we investigated the effects of leakage on machine learning models in two common neuroimaging modalities, functional and structural connectomes. Using over 400 different pipelines spanning four large datasets and three phenotypes, we evaluated five forms of leakage fitting into three broad categories: feature selection, covariate correction, and lack of independence between subjects. As expected, leakage via feature selection and repeated subjects drastically inflated prediction performance. Notably, other forms of leakage had only minor effects (e.g., leaky site correction) or even decreased prediction performance (e.g., leaky covariate regression). In some cases, leakage affected not only prediction performance, but also model coefficients, and thus neurobiological interpretations. Finally, we found that predictive models using small datasets were more sensitive to leakage. Overall, our results illustrate the variable effects of leakage on prediction pipelines and underscore the importance of avoiding data leakage to improve the validity and reproducibility of predictive modeling.

Exploring individual fixel-based white matter abnormalities in epilepsy

Diffusion MRI has provided insight into the widespread structural connectivity changes that characterize epilepsies. Although syndrome-specific white matter abnormalities have been demonstrated, studies to date have predominantly relied on statistical comparisons between patient and control groups. For diffusion MRI techniques to be of clinical value, they should be able to detect white matter microstructural changes in individual patients. In this study, we apply an individualized approach to a technique known as fixel-based analysis, to examine fibre-tract-specific abnormalities in individuals with epilepsy. We explore the potential clinical value of this individualized fixel-based approach in epilepsy patients with differing syndromic diagnoses. Diffusion MRI data from 90 neurologically healthy control participants and 10 patients with epilepsy (temporal lobe epilepsy, progressive myoclonus epilepsy, and Dravet Syndrome, malformations of cortical development) were included in this study. Measures of fibre density and cross-section were extracted for all participants across brain white matter fixels, and mean values were computed within select tracts-of-interest. Scanner harmonized and normalized data were then used to compute Z-scores for individual patients with epilepsy. White matter abnormalities were observed in distinct patterns in individual patients with epilepsy, both at the tract and fixel level. For patients with specific epilepsy syndromes, the detected white matter abnormalities were in line with expected syndrome-specific clinical phenotypes. In patients with lesional epilepsies (e.g. hippocampal sclerosis, periventricular nodular heterotopia, and bottom-of-sulcus dysplasia), white matter abnormalities were spatially concordant with lesion location. This proof-of-principle study demonstrates the clinical potential of translating advanced diffusion MRI methodology to individual-patient-level use in epilepsy. This technique could be useful both in aiding diagnosis of specific epilepsy syndromes, and in localizing structural abnormalities, and is readily amenable to other neurological disorders. We have included code and data for this study so that individualized white matter changes can be explored robustly in larger cohorts in future work.

Recalibrating single-study effect sizes using hierarchical Bayesian models

Introduction

There are growing concerns about commonly inflated effect sizes in small neuroimaging studies, yet no study has addressed recalibrating effect size estimates for small samples. To tackle this issue, we propose a hierarchical Bayesian model to adjust the magnitude of single-study effect sizes while incorporating a tailored estimation of sampling variance.

Methods

We estimated the effect sizes of case-control differences on brain structural features between individuals who were dependent on alcohol, nicotine, cocaine, methamphetamine, or cannabis and non-dependent participants for 21 individual studies (Total cases: 903; Total controls: 996). Then, the study-specific effect sizes were modeled using a hierarchical Bayesian approach in which the parameters of the study-specific effect size distributions were sampled from a higher-order overarching distribution. The posterior distribution of the overarching and study-specific parameters was approximated using the Gibbs sampling method.

Results

The results showed shrinkage of the posterior distribution of the study-specific estimates toward the overarching estimates given the original effect sizes observed in individual studies. Differences between the original effect sizes (i.e., Cohen's d) and the point estimate of the posterior distribution ranged from 0 to 0.97. The magnitude of adjustment was negatively correlated with the sample size (r = -0.27, p < 0.001) and positively correlated with empirically estimated sampling variance (r = 0.40, p < 0.001), suggesting studies with smaller samples and larger sampling variance tended to have greater adjustments.

Discussion

Our findings demonstrate the utility of the hierarchical Bayesian model in recalibrating single-study effect sizes using information from similar studies. This suggests that Bayesian utilization of existing knowledge can be an effective alternative approach to improve the effect size estimation in individual studies, particularly for those with smaller samples.

Personalized functional brain network topography is associated with individual differences in youth cognition

Individual differences in cognition during childhood are associated with important social, physical, and mental health outcomes in adolescence and adulthood. Given that cortical surface arealization during development reflects the brain's functional prioritization, quantifying variation in the topography of functional brain networks across the developing cortex may provide insight regarding individual differences in cognition. We test this idea by defining personalized functional networks (PFNs) that account for interindividual heterogeneity in functional brain network topography in 9-10 year olds from the Adolescent Brain Cognitive Development℠ Study. Across matched discovery (n = 3525) and replication (n = 3447) samples, the total cortical representation of fronto-parietal PFNs positively correlates with general cognition. Cross-validated ridge regressions trained on PFN topography predict cognition in unseen data across domains, with prediction accuracy increasing along the cortex's sensorimotor-association organizational axis. These results establish that functional network topography heterogeneity is associated with individual differences in cognition before the critical transition into adolescence.

BrainAGE Estimation: Influence of Field Strength, Voxel Size, Race, and Ethnicity

The BrainAGE method is used to estimate biological brain age using structural neuroimaging. However, the stability of the model across different scan parameters and races/ethnicities has not been thoroughly investigated. Estimated brain age was compared within- and across- MRI field strength and across voxel sizes. Estimated brain age gap (BAG) was compared across demographically matched groups of different self-reported races and ethnicities in ADNI and IMAS cohorts. Longitudinal ComBat was used to correct for potential scanner effects. The brain age method was stable within field strength, but less stable across different field strengths. The method was stable across voxel sizes. There was a significant difference in BAG between races, but not ethnicities. Correction procedures are suggested to eliminate variation across scanner field strength while maintaining accurate brain age estimation. Further studies are warranted to determine the factors contributing to racial differences in BAG.

Effect of data harmonization of multicentric dataset in ASD/TD classification

Machine Learning (ML) is nowadays an essential tool in the analysis of Magnetic Resonance Imaging (MRI) data, in particular in the identification of brain correlates in neurological and neurodevelopmental disorders. ML requires datasets of appropriate size for training, which in neuroimaging are typically obtained collecting data from multiple acquisition centers. However, analyzing large multicentric datasets can introduce bias due to differences between acquisition centers. ComBat harmonization is commonly used to address batch effects, but it can lead to data leakage when the entire dataset is used to estimate model parameters. In this study, structural and functional MRI data from the Autism Brain Imaging Data Exchange (ABIDE) collection were used to classify subjects with Autism Spectrum Disorders (ASD) compared to Typical Developing controls (TD). We compared the classical approach (external harmonization) in which harmonization is performed before train/test split, with an harmonization calculated only on the train set (internal harmonization), and with the dataset with no harmonization. The results showed that harmonization using the whole dataset achieved higher discrimination performance, while non-harmonized data and harmonization using only the train set showed similar results, for both structural and connectivity features. We also showed that the higher performances of the external harmonization are not due to larger size of the sample for the estimation of the model and hence these improved performance with the entire dataset may be ascribed to data leakage. In order to prevent this leakage, it is recommended to define the harmonization model solely using the train set.

Enhancing histopathological image classification of invasive ductal carcinoma using hybrid harmonization techniques

This study aims to develop a robust pipeline for classifying invasive ductal carcinomas and benign tumors in histopathological images, addressing variability within and between centers. We specifically tackle the challenge of detecting atypical data and variability between common clusters within the same database. Our feature engineering-based pipeline comprises a feature extraction step, followed by multiple harmonization techniques to rectify intra- and inter-center batch effects resulting from image acquisition variability and diverse patient clinical characteristics. These harmonization steps facilitate the construction of more robust and efficient models. We assess the proposed pipeline's performance on two public breast cancer databases, BreaKHIS and IDCDB, utilizing recall, precision, and accuracy metrics. Our pipeline outperforms recent models, achieving 90-95% accuracy in classifying benign and malignant tumors. We demonstrate the advantage of harmonization for classifying patches from different databases. Our top model scored 94.7% for IDCDB and 95.2% for BreaKHis, surpassing existing feature engineering-based models (92.1% for IDCDB and 87.7% for BreaKHIS) and attaining comparable performance to deep learning models. The proposed feature-engineering-based pipeline effectively classifies malignant and benign tumors while addressing variability within and between centers through the incorporation of various harmonization techniques. Our findings reveal that harmonizing variabilities between patches from different batches directly impacts the learning and testing performance of classification models. This pipeline has the potential to enhance breast cancer diagnosis and treatment and may be applicable to other diseases.

Traveling Subject-Informed Harmonization Increases Reliability of Brain Diffusion Tensor and Neurite Mapping

Diffusion-weighted magnetic resonance imaging (dMRI) of brain has helped elucidate the microstructural changes of psychiatric and neurodegenerative disorders. Inconsistency between MRI models has hampered clinical application of dMRI-based metrics. Using harmonized dMRI data of 300 scans from 69 traveling subjects (TS) scanning the same individuals at multiple conditions with 13 MRI models and 2 protocols, the widely-used metrics such as diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI) were evaluated before and after harmonization with a combined association test (ComBat) or TS-based general linear model (TS-GLM). Results showed that both ComBat and TS-GLM significantly reduced the effects of the MRI site, model, and protocol for diffusion metrics while maintaining the intersubject biological effects. The harmonization power of TS-GLM based on TS data model is more powerful than that of ComBat. In conclusion, our research demonstrated that although ComBat and TS-GLM harmonization approaches were effective at reducing the scanner effects of the site, model, and protocol for DTI and NODDI metrics in WM, they exhibited high retainability of biological effects. Therefore, we suggest that, after harmonizing DTI and NODDI metrics, a multisite study with large cohorts can accurately detect small pathological changes by retaining pathological effects.

Harmonization of multi-site diffusion tensor imaging data for cervical and thoracic spinal cord at 1.5 T and 3 T using longitudinal ComBat

MRI scanner hardware, field strengths, and sequence parameters are major variables in diffusion studies of the spinal cord. Reliability between scanners is not well known, particularly for the thoracic cord. DTI data was collected for the entire cervical and thoracic spinal cord in thirty healthy adult subjects with different MR vendors and field strengths. DTI metrics were extracted and averaged for all slices within each vertebral level. Metrics were examined for variability and then harmonized using longitudinal ComBat (longComBat). Four scanners were used: Siemens 3 T Prisma, Siemens 1.5 T Avanto, Philips 3 T Ingenia, Philips 1.5 T Achieva. Average full cord diffusion values/standard deviation for all subjects and scanners were FA: 0.63, σ = 0.10, MD: 1.11, σ = 0.12 × 10-3 mm2/s, AD: 1.98, σ = 0.55 × 10-3 mm2/s, RD: 0.67, σ = 0.31 × 10-3 mm2/s. FA metrics averaged for all subjects by level were relatively consistent across scanners, but large variability was found in diffusivity measures. Coefficients of variation were lowest in the cervical region, and relatively lower for FA than diffusivity measures. Harmonized metrics showed greatly improved agreement between scanners. Variability in DTI of the spinal cord arises from scanner hardware differences, pulse sequence differences, physiological motion, and subject compliance. The use of longComBat resulted in large improvement in agreement of all DTI metrics between scanners. This study shows the importance of harmonization of diffusion data in the spinal cord and potential for longitudinal and multisite clinical research and clinical trials.

A common [18F]-FDG PET radiomic signature to predict survival in patients with HPV-induced cancers

Locally advanced cervical cancer (LACC) and anal and oropharyngeal squamous cell carcinoma (ASCC and OPSCC) are mostly caused by oncogenic human papillomaviruses (HPV). In this paper, we developed machine learning (ML) models based on clinical, biological, and radiomic features extracted from pre-treatment fluorine-18-fluorodeoxyglucose positron emission tomography ([18F]-FDG PET) images to predict the survival of patients with HPV-induced cancers. For this purpose, cohorts from five institutions were used: two cohorts of patients treated for LACC including 104 patients from Gustave Roussy Campus Cancer (Center 1) and 90 patients from Leeds Teaching Hospitals NHS Trust (Center 2), two datasets of patients treated for ASCC composed of 66 patients from Institut du Cancer de Montpellier (Center 3) and 67 patients from Oslo University Hospital (Center 4), and one dataset of 45 OPSCC patients from the University Hospital of Zurich (Center 5). Radiomic features were extracted from baseline [18F]-FDG PET images. The ComBat technique was applied to mitigate intra-scanner variability. A modified consensus nested cross-validation for feature selection and hyperparameter tuning was applied on four ML models to predict progression-free survival (PFS) and overall survival (OS) using harmonized imaging features and/or clinical and biological variables as inputs. Each model was trained and optimized on Center 1 and Center 3 cohorts and tested on Center 2, Center 4, and Center 5 cohorts. The radiomic-based CoxNet model achieved C-index values of 0.75 and 0.78 for PFS and 0.76, 0.74, and 0.75 for OS on the test sets. Radiomic feature-based models had superior performance compared to the bioclinical ones, and combining radiomic and bioclinical variables did not improve the performances. Metabolic tumor volume (MTV)-based models obtained lower C-index values for a majority of the tested configurations but quite equivalent performance in terms of time-dependent AUCs (td-AUC). The results demonstrate the possibility of identifying common PET-based image signatures for predicting the response of patients with induced HPV pathology, validated on multi-center multiconstructor data.

Complementary structural and functional abnormalities to localise epileptogenic tissue

BACKGROUND

When investigating suitability for epilepsy surgery, people with drug-refractory focal epilepsy may have intracranial EEG (iEEG) electrodes implanted to localise seizure onset. Diffusion-weighted magnetic resonance imaging (dMRI) may be acquired to identify key white matter tracts for surgical avoidance. Here, we investigate whether structural connectivity abnormalities, inferred from dMRI, may be used in conjunction with functional iEEG abnormalities to aid localisation of the epileptogenic zone (EZ), improving surgical outcomes in epilepsy.

METHODS

We retrospectively investigated data from 43 patients (42% female) with epilepsy who had surgery following iEEG. Twenty-five patients (58%) were free from disabling seizures (ILAE 1 or 2) at one year. Interictal iEEG functional, and dMRI structural connectivity abnormalities were quantified by comparison to a normative map and healthy controls. We explored whether the resection of maximal abnormalities related to improved surgical outcomes, in both modalities individually and concurrently. Additionally, we suggest how connectivity abnormalities may inform the placement of iEEG electrodes pre-surgically using a patient case study.

FINDINGS

Seizure freedom was 15 times more likely in patients with resection of maximal connectivity and iEEG abnormalities (p = 0.008). Both modalities separately distinguished patient surgical outcome groups and when used simultaneously, a decision tree correctly separated 36 of 43 (84%) patients.

INTERPRETATION

Our results suggest that both connectivity and iEEG abnormalities may localise epileptogenic tissue, and that these two modalities may provide complementary information in pre-surgical evaluations.

FUNDING

This research was funded by UKRI, CDT in Cloud Computing for Big Data, NIH, MRC, Wellcome Trust and Epilepsy Research UK.

Large-scale Normative Modeling of Brain Microstructure

Normative models of brain metrics based on large populations are extremely valuable for detecting brain abnormalities in patients with dementia, psychiatric, or developmental conditions. Here we present the first large-scale normative model of the brain's white matter (WM) microstructure derived from 18 international diffusion MRI (dMRI) datasets covering almost the entire lifespan (totaling N=51,830 individuals; age: 3-80 years). We extracted regional diffusion tensor imaging (DTI) metrics using a standardized analysis and quality control protocol, and used Hierarchical Bayesian Regression (HBR) to model the statistical distribution of derived WM metrics as a function of age and sex, while modeling the site effect. HBR overcomes known weaknesses of some data harmonization methods that simply scale and shift residual distributions at each site. To illustrate the method, we applied it to detect and visualize profiles of WM microstructural deviations in cohorts of patients with Alzheimer's disease, mild cognitive impairment, Parkinson's disease and in carriers of 22q11.2 copy number variants, a rare neurogenetic condition that confers increased risk for psychosis. The resulting large-scale model offers a common reference to identify disease effects in individuals or groups, as well as to compare disorders and discover factors that influence these abnormalities.

A Weakly Supervised Gradient Attribution Constraint for Interpretable Classification and Anomaly Detection

The lack of interpretability of deep learning reduces understanding of what happens when a network does not work as expected and hinders its use in critical fields like medicine, which require transparency of decisions. For example, a healthy vs pathological classification model should rely on radiological signs and not on some training dataset biases. Several post-hoc models have been proposed to explain the decision of a trained network. However, they are very seldom used to enforce interpretability during training and none in accordance with the classification. In this paper, we propose a new weakly supervised method for both interpretable healthy vs pathological classification and anomaly detection. A new loss function is added to a standard classification model to constrain each voxel of healthy images to drive the network decision towards the healthy class according to gradient-based attributions. This constraint reveals pathological structures for patient images, allowing their unsupervised segmentation. Moreover, we advocate both theoretically and experimentally, that constrained training with the simple Gradient attribution is similar to constraints with the heavier Expected Gradient, consequently reducing the computational cost. We also propose a combination of attributions during the constrained training making the model robust to the attribution choice at inference. Our proposition was evaluated on two brain pathologies: tumors and multiple sclerosis. This new constraint provides a more relevant classification, with a more pathology-driven decision. For anomaly detection, the proposed method outperforms state-of-the-art especially on difficult multiple sclerosis lesions segmentation task with a 15 points Dice improvement.

Visuo-vestibular and cognitive connections of the vestibular neuromatrix are conserved across age and injury populations

BACKGROUND AND PURPOSE

Given the prevalence of vestibular dysfunction in pediatric concussion, there is a need to better understand pathophysiological disruptions within vestibular and associated cognitive, affective, and sensory-integrative networks. Although current research leverages established intrinsic connectivity networks, these are nonspecific for vestibular function, suggesting that a pathologically guided approach is warranted. The purpose of this study was to evaluate the generalizability of the previously identified "vestibular neuromatrix" in adults with and without postconcussive vestibular dysfunction to young athletes aged 14-17.

METHODS

This retrospective study leveraged resting-state functional MRI data from two sites. Site A included adults with diagnosed postconcussive vestibular impairment and healthy adult controls and Site B consisted of young athletes with preseason, postconcussion, and postseason time points (prospective longitudinal data). Adjacency matrices were generated from preprocessed resting-state data from each sample and assessed for overlap and network structure in MATLAB.

RESULTS

Analyses indicated the presence of a conserved "core" network of vestibular regions as well as areas subserving visual, spatial, and attentional processing. Other vestibular connections were also conserved across samples but were not linked to the "core" subnetwork by regions of interest included in this study.

CONCLUSIONS

Our results suggest that connections between central vestibular, visuospatial, and known intrinsic connectivity networks are conserved across adult and pediatric participants with and without concussion, evincing the significance of this expanded, vestibular-associated network. Our findings thus support this network as a workable model for investigation in future studies of dysfunction in young athlete populations.

Abnormal individualized peak functional connectivity toward potential repetitive transcranial magnetic stimulation treatment of autism spectrum disorder

Functional connectivity (FC) derived from resting-state functional magnetic resonance imaging has been widely applied to guide precise repetitive transcranial magnetic stimulation (rTMS). The left, right, and bilateral dorsolateral prefrontal cortices (DLPFC) have been used as rTMS treatment target regions for autism spectrum disorder (ASD), albeit with moderate efficacy. Thus, we aimed to develop an individualized localization method for rTMS treatment of ASD. We included 266 male ASDs and 297 male typically-developed controls (TDCs) from the Autism Brain Imaging Data Exchange Dataset. The nucleus accumbens (NAc) was regarded as a promising effective region, which was used as a seed and individualized peak FC strength in the DLPFC was compared between ASD and TDC. Correlation analysis was conducted between individualized peak FC strength and symptoms in ASD. We also investigated the spatial distribution of individualized peak FC locations in the DLPFC and conducted voxel-wise analysis to compare NAc-based FC between the two groups. ASD showed stronger peak FC in the right DLPFC related to TDC (Cohen's d = -.19, 95% CI: -0.36 to -0.03, t = -2.30, p = .02). Moreover, negative correlation was found between the peak FC strength in the right DLPFC and Autism Diagnostic Observation Schedule (ADOS) scores, which assessed both the social communication and interaction (r = -.147, p = .04, uncorrected significant), and stereotyped behaviors and restricted interests (r = -.198, p = .02, corrected significant). Peak FC locations varied substantially across participants. No significant differences in NAc-based FC in the DLPFC were found in the voxel-wise comparison. Our study supports the use of individualized peak FC-guided precise rTMS treatment of male ASD. Moreover, stimulating the right DLPFC might alleviate core symptoms of ASD.

Dual-Centre Harmonised Multimodal Positron Emission Tomography/Computed Tomography Image Radiomic Features and Machine Learning Algorithms for Non-small Cell Lung Cancer Histopathological Subtype Phenotype Decoding

AIMS

We aimed to build radiomic models for classifying non-small cell lung cancer (NSCLC) histopathological subtypes through a dual-centre dataset and comprehensively evaluate the effect of ComBat harmonisation on the performance of single- and multimodality radiomic models.

MATERIALS AND METHODS

A public dataset of NSCLC patients from two independent centres was used. Two image fusion methods, namely guided filtering-based fusion and image fusion based on visual saliency map and weighted least square optimisation, were used. Radiomic features were extracted from each scan, including first-order, texture and moment-invariant features. Subsequently, ComBat harmonisation was applied to the extracted features from computed tomography (CT), positron emission tomography (PET) and fused images to correct the centre effect. For feature selection, least absolute shrinkage and selection operator (Lasso) and recursive feature elimination (RFE) were investigated. For machine learning, logistic regression (LR), support vector machine (SVM) and AdaBoost were evaluated for classifying NSCLC subtypes. Training and evaluation of the models were carried out in a robust framework to offset plausible errors and performance was reported using area under the curve, balanced accuracy, sensitivity and specificity before and after harmonisation. N-way ANOVA was used to assess the effect of different factors on the performance of the models.

RESULTS

Support vector machine fed with selected features by recursive feature elimination from a harmonised PET feature set achieved the highest performance (area under the curve = 0.82) in classifying NSCLC histopathological subtypes. Although the performance of the models did not significantly improve for CT images after harmonisation, the performance of PET and guided filtering-based fusion feature signatures significantly improved for almost all models. Although the selection of the image modality and feature selection methods was effective on the performance of the model (ANOVA P-values <0.001), machine learning and harmonisation did not change the performance significantly (ANOVA P-values = 0.839 and 0.292, respectively).

CONCLUSION

This study confirmed the potential of radiomic analysis on PET, CT and hybrid images for histopathological classification of NSCLC subtypes.

Diffusion in the corpus callosum predicts persistence of clinical symptoms after mild traumatic brain injury, a multi-scanner study

Background

Mild traumatic brain injuries (mTBIs) comprise 80% of all TBI, but conventional MRI techniques are often insensitive to the subtle changes and injuries produced in a concussion. Diffusion tensor imaging (DTI) is one of the most sensitive MRI techniques for mTBI studies with outcome and symptom associations described. The corpus callosum (CC) is one of the most studied fiber tracts in TBI and mTBI, but the comprehensive post-mTBI symptom relationship has not fully been explored.

Methods

This is a retrospective observational study of how quantitative DTI data of the CC and its sub-regions may relate to clinical presentation of symptoms and timing of resolution of symptoms in patients diagnosed with uncomplicated mTBI. DTI and clinical data were obtained retrospectively from 446 (mean age 42 years, range 13-82) civilian patients. From patient medical charts, presentation of the following common post-concussive symptoms was noted: headache, balance issues, cognitive deficits, fatigue, anxiety, depression, and emotional lability. Also recorded was the time between injury and a visit to the physician when improvement or resolution of a particular symptom was reported. FA values from the total CC and 3 subregions of the CC (genu or anterior, mid body, and splenium or posterior) were obtained from hand tracing on the Olea Sphere v3.0 SP12 free-standing workstation. DTI data was obtained from 8 different 3T MRI scanners and harmonized via ComBat harmonization. The statistical models used to explore the association between regional Fractional Anisotropy (FA) values and symptom presentation and time to symptom resolution were logistic regression and interval-censored semi-parametric Cox proportional hazard models, respectively. Subgroups related to age and timing of first scan were also analyzed.

Results

Patients with the highest FA in the total CC (p = 0.01), anterior CC (p < 0.01), and mid-body CC (p = 0.03), but not the posterior CC (p = 0.91) recovered faster from post-concussive cognitive deficits. Patients with the highest FA in the posterior CC recovered faster from depression (p = 0.04) and emotional lability (p = 0.01). There was no evidence that FA in the CC or any of its sub-regions was associated with symptom presentation or with time to resolution of headache, balance issues, fatigue, or anxiety. Patients with mTBI under 40 had higher FA in the CC and the anterior and mid-body subregions (but not the posterior subregion: p = 1.00) compared to patients 40 or over (p ≤ 0.01). There was no evidence for differences in symptom presentation based on loss of consciousness (LOC) or sex (p ≥ 0.18).

Conclusion

This study suggests that FA of the CC has diagnostic and prognostic value for clinical assessment of mTBI in a large diverse civilian population, particularly in patients with cognitive symptoms.