Intensity warping for multisite MRI harmonization.
Ontology highlight
ABSTRACT:
SUBMITTER: Wrobel J
PROVIDER: S-EPMC8412236 | biostudies-literature | 2020 Dec
REPOSITORIES: biostudies-literature
Publications
Intensity warping for multisite MRI harmonization.
Wrobel J J Martin M L ML Bakshi R R Calabresi P A PA Elliot M M Roalf D D Gur R C RC Gur R E RE Henry R G RG Nair G G Oh J J Papinutto N N Pelletier D D Reich D S DS Rooney W D WD Satterthwaite T D TD Stern W W Prabhakaran K K Sicotte N L NL Shinohara R T RT Goldsmith J J
NeuroImage 20200814
In multisite neuroimaging studies there is often unwanted technical variation across scanners and sites. These "scanner effects" can hinder detection of biological features of interest, produce inconsistent results, and lead to spurious associations. We propose mica (multisite image harmonization by cumulative distribution function alignment), a tool to harmonize images taken on different scanners by identifying and removing within-subject scanner effects. Our goals in the present study were to ...[more]
Similar Datasets
Project description:When data is pooled across multiple sites, the extracted features are confounded by site effects. Harmonization methods attempt to correct these site effects while preserving the biological variability within the features. However, little is known about the sample size requirement for effectively learning the harmonization parameters and their relationship with the increasing number of sites. In this study, we performed experiments to find the minimum sample size required to achieve multisite harmonization (using neuroHarmonize) using volumetric and surface features by leveraging the concept of learning curves. Our first two experiments show that site-effects are effectively removed in a univariate and multivariate manner; however, it is essential to regress the effect of covariates from the harmonized data additionally. Our following two experiments with actual and simulated data showed that the minimum sample size required for achieving harmonization grows with the increasing average Mahalanobis distances between the sites and their reference distribution. We conclude by positing a general framework to understand the site effects using the Mahalanobis distance. Further, we provide insights on the various factors in a cross-validation design to achieve optimal inter-site harmonization.
| S-EPMC7615107 | biostudies-literature
Project description:The analysis of large, multisite neuroimaging datasets provides a promising means for robust characterization of brain networks that can reduce false positives and improve reproducibility. However, the use of different MRI scanners introduces variability to the data. Managing those sources of variability is increasingly important for the generation of accurate group-level inferences. ComBat is one of the most promising tools for multisite (multiscanner) harmonization of structural neuroimaging data, but no study has examined its application to graph theory metrics derived from the structural brain connectome. The present work evaluates the use of ComBat for multisite harmonization in the context of structural network analysis of diffusion-weighted scans from the Advancing Concussion Assessment in Pediatrics (A-CAP) study. Scans were acquired on six different scanners from 484 children aged 8.00-16.99 years [Mean = 12.37 ± 2.34 years; 289 (59.7%) Male] ~10 days following mild traumatic brain injury (n = 313) or orthopedic injury (n = 171). Whole brain deterministic diffusion tensor tractography was conducted and used to construct a 90 x 90 weighted (average fractional anisotropy) adjacency matrix for each scan. ComBat harmonization was applied separately at one of two different stages during data processing, either on the (i) weighted adjacency matrices (matrix harmonization) or (ii) global network metrics derived using unharmonized weighted adjacency matrices (parameter harmonization). Global network metrics based on unharmonized adjacency matrices and each harmonization approach were derived. Robust scanner effects were found for unharmonized metrics. Some scanner effects remained significant for matrix harmonized metrics, but effect sizes were less robust. Parameter harmonized metrics did not differ by scanner. Intraclass correlations (ICC) indicated good to excellent within-scanner consistency between metrics calculated before and after both harmonization approaches. Age correlated with unharmonized network metrics, but was more strongly correlated with network metrics based on both harmonization approaches. Parameter harmonization successfully controlled for scanner variability while preserving network topology and connectivity weights, indicating that harmonization of global network parameters based on unharmonized adjacency matrices may provide optimal results. The current work supports the use of ComBat for removing multiscanner effects on global network topology.
| S-EPMC9247315 | biostudies-literature
Project description:Cognitive deficits are among the best predictors of real-world functioning in schizophrenia. However, our understanding of how cognitive deficits relate to neuropathology and clinical presentation over the disease lifespan is limited. Here, we combine multi-site, harmonized cognitive, imaging, demographic, and clinical data from over 900 individuals to characterize a) cognitive deficits across the schizophrenia lifespan and b) the association between cognitive deficits, clinical presentation, and white matter (WM) microstructure. Multimodal harmonization was accomplished using T-scores for cognitive data, previously reported standardization methods for demographic and clinical data, and an established harmonization method for imaging data. We applied t-tests and correlation analysis to describe cognitive deficits in individuals with schizophrenia. We then calculated whole-brain WM fractional anisotropy (FA) and utilized regression-mediation analyses to model the association between diagnosis, FA, and cognitive deficits. We observed pronounced cognitive deficits in individuals with schizophrenia (p < 0.006), associated with more positive symptoms and medication dosage. Regression-mediation analyses showed that WM microstructure mediated the association between schizophrenia and language/processing speed/working memory/non-verbal memory. In addition, processing speed mediated the influence of diagnosis and WM microstructure on the other cognitive domains. Our study highlights the critical role of cognitive deficits in schizophrenia. We further show that WM is crucial when trying to understand the role of cognitive deficits, given that it explains the association between schizophrenia and cognitive deficits (directly and via processing speed).
| S-EPMC10538303 | biostudies-literature
Project description:Increasing numbers of 7 T (7 T) magnetic resonance imaging (MRI) scanners are in research and clinical use. 7 T MRI can increase the scanning speed, spatial resolution and contrast-to-noise-ratio of many neuroimaging protocols, but technical challenges in implementation have been addressed in a variety of ways across sites. In order to facilitate multi-centre studies and ensure consistency of findings across sites, it is desirable that 7 T MRI sites implement common high-quality neuroimaging protocols that can accommodate different scanner models and software versions. With the installation of several new 7 T MRI scanners in the United Kingdom, the UK7T Network was established with an aim to create a set of harmonized structural and functional neuroimaging sequences and protocols. The Network currently includes five sites, which use three different scanner platforms, provided by two different vendors. Here we describe the harmonization of functional and anatomical imaging protocols across the three different scanner models, detailing the necessary changes to pulse sequences and reconstruction methods. The harmonized sequences are fully described, along with implementation details. Example datasets acquired from the same subject on all Network scanners are made available. Based on these data, an evaluation of the harmonization is provided. In addition, the implementation and validation of a common system calibration process is described.
| S-EPMC7212005 | biostudies-literature
Project description:Multi-site brain magnetic resonance imaging (MRI) has been widely used in clinical and research domains, but usually is sensitive to non-biological variations caused by site effects (e.g., field strengths and scanning protocols). Several retrospective data harmonization methods have shown promising results in removing these non-biological variations at feature or whole-image level. Most existing image-level harmonization methods are implemented through generative adversarial networks, which are generally computationally expensive and generalize poorly on independent data. To this end, this paper proposes a disentangled latent energy-based style translation (DLEST) framework for image-level structural MRI harmonization. Specifically, DLEST disentangles site-invariant image generation and site-specific style translation via a latent autoencoder and an energy-based model. The autoencoder learns to encode images into low-dimensional latent space, and generates faithful images from latent codes. The energy-based model is placed in between the encoding and generation steps, facilitating style translation from a source domain to a target domain implicitly. This allows highly generalizable image generation and efficient style translation through the latent space. We train our model on 4,092 T1-weighted MRIs in 3 tasks: histogram comparison, acquisition site classification, and brain tissue segmentation. Qualitative and quantitative results demonstrate the superiority of our approach, which generally outperforms several state-of-the-art methods.
| S-EPMC10883146 | biostudies-literature
Project description:ObjectiveSimultaneous PET/MRIs vary in their quantitative PET performance due to inherent differences in the physical systems and differences in the image reconstruction implementation. This variability in quantitative accuracy confounds the ability to meaningfully combine and compare data across scanners. In this work, we define image reconstruction parameters that lead to comparable contrast recovery curves across simultaneous PET/MRI systems.MethodThe NEMA NU-2 image quality phantom was imaged on one GE Signa and on one Siemens mMR PET/MRI scanner. The phantom was imaged at 9.7:1 contrast with standard spheres (diameter 10, 13, 17, 22, 28, 37 mm) and with custom spheres (diameter: 8.5, 11.5, 15, 25, 32.5, 44 mm) using a standardized methodology. Analysis was performed on a 30 min listmode data acquisition and on 6 realizations of 5 min from the listmode data. Images were reconstructed with the manufacturer provided iterative image reconstruction algorithms with and without point spread function (PSF) modeling. For both scanners, a post-reconstruction Gaussian filter of 3-7 mm in steps of 1 mm was applied. Attenuation correction was provided from a scaled computed tomography (CT) image of the phantom registered to the MR-based attenuation images and verified to align on the non-attenuation corrected PET images. For each of these image reconstruction parameter sets, contrast recovery coefficients (CRCs) were determined for the SUVmean, SUVmax and SUVpeak for each sphere. A hybrid metric combining the root-mean-squared discrepancy (RMSD) and the absolute CRC values was used to simultaneously optimize for best match in CRC between the two scanners while simultaneously weighting toward higher resolution reconstructions. The image reconstruction parameter set was identified as the best candidate reconstruction for each vendor for harmonized PET image reconstruction.ResultsThe range of clinically relevant image reconstruction parameters demonstrated widely different quantitative performance across cameras. The best match of CRC curves was obtained at the lowest RMSD values with: for CRCmean, 2 iterations-7 mm filter on the GE Signa and 4 iterations-6 mm filter on the Siemens mMR, for CRCmax, 4 iterations-6 mm filter on the GE Signa, 4 iterations-5 mm filter on the Siemens mMR and for CRCpeak, 4 iterations-7 mm filter with PSF on the GE Signa and 4 iterations-7 mm filter on the Siemens mMR. Over all reconstructions, the RMSD between CRCs was 1.8%, 3.6% and 2.9% for CRC mean, max and peak, respectively. The solution of 2 iterations-3 mm on the GE Signa and 4 iterations-3 mm on Siemens mMR, both with PSF, led to simultaneous harmonization and with high CRC and low RMSD for CRC mean, max and peak with RMSD values of 2.8%, 5.8% and 3.2%, respectively.ConclusionsFor two commercially available PET/MRI scanners, user-selectable parameters that control iterative updates, image smoothing and PSF modeling provide a range of contrast recovery curves that allow harmonization in harmonization strategies of optimal match in CRC or high CRC values. This work demonstrates that nearly identical CRC curves can be obtained on different commercially available scanners by selecting appropriate image reconstruction parameters.
| S-EPMC8571452 | biostudies-literature
Project description:Classification of disease and healthy volunteer cohorts provides a useful clinical alternative to traditional group statistics due to individualized, personalized predictions. Classifiers for neurodegenerative disease can be trained on structural MRI morphometry, but require large multi-scanner datasets, introducing confounding batch effects. We test ComBat, a common harmonization model, in an example application to classify subjects with Parkinson's disease from healthy volunteers and identify common pitfalls, including data leakage. We used a multi-dataset cohort of 372 subjects (216 with Parkinson's disease, 156 healthy volunteers) from 11 identified scanners. We extracted both FreeSurfer and the determinant of Jacobian morphometry to compare single-scanner and multi-scanner classification pipelines. We confirm the presence of batch effects by running single scanner classifiers which could achieve wildly divergent AUCs on scanner-specific datasets (mean:0.651 ± 0.144). Multi-scanner classifiers that considered neurobiological batch effects between sites could easily achieve a test AUC of 0.902, though pipelines that prevented data leakage could only achieve a test AUC of 0.550. We conclude that batch effects remain a major issue for classification problems, such that even impressive single-scanner classifiers are unlikely to generalize to multiple scanners, and that solving for batch effects in a classifier problem must avoid circularity and reporting overly optimistic results.
| S-EPMC11678312 | biostudies-literature
Project description:Diffusion magnetic resonance imaging can indirectly infer the microstructure of tissues and provide metrics subject to normal variability in a population. Potentially abnormal values may yield essential information to support analysis of controls and patients cohorts, but subtle confounds could be mistaken for purely biologically driven variations amongst subjects. In this work, we propose a new harmonization algorithm based on adaptive dictionary learning to mitigate the unwanted variability caused by different scanner hardware while preserving the natural biological variability of the data. Our harmonization algorithm does not require paired training data sets, nor spatial registration or matching spatial resolution. Overcomplete dictionaries are learned iteratively from all data sets at the same time with an adaptive regularization criterion, removing variability attributable to the scanners in the process. The obtained mapping is applied directly in the native space of each subject toward a scanner-space. The method is evaluated with a public database which consists of two different protocols acquired on three different scanners. Results show that the effect size of the four studied diffusion metrics is preserved while removing variability attributable to the scanner. Experiments with alterations using a free water compartment, which is not simulated in the training data, shows that the modifications applied to the diffusion weighted images are preserved in the diffusion metrics after harmonization, while still reducing global variability at the same time. The algorithm could help multicenter studies pooling their data by removing scanner specific confounds, and increase statistical power in the process.
| S-EPMC7555079 | biostudies-literature
Project description:Neuroimaging studies often lack reproducibility, one of the cardinal features of the scientific method. Multisite collaboration initiatives increase sample size and limit methodological flexibility, therefore providing the foundation for increased statistical power and generalizable results. However, multisite collaborative initiatives are inherently limited by hardware, software, and pulse and sequence design heterogeneities of both clinical and preclinical MRI scanners and the lack of benchmark for acquisition protocols, data analysis, and data sharing. We present the overarching vision that yielded to the constitution of RIN-Neuroimaging Network, a national consortium dedicated to identifying disease and subject-specific in-vivo neuroimaging biomarkers of diverse neurological and neuropsychiatric conditions. This ambitious goal needs efforts toward increasing the diagnostic and prognostic power of advanced MRI data. To this aim, 23 Italian Scientific Institutes of Hospitalization and Care (IRCCS), with technological and clinical specialization in the neurological and neuroimaging field, have gathered together. Each IRCCS is equipped with high- or ultra-high field MRI scanners (i.e., ≥3T) for clinical or preclinical research or has established expertise in MRI data analysis and infrastructure. The actions of this Network were defined across several work packages (WP). A clinical work package (WP1) defined the guidelines for a minimum standard clinical qualitative MRI assessment for the main neurological diseases. Two neuroimaging technical work packages (WP2 and WP3, for clinical and preclinical scanners) established Standard Operative Procedures for quality controls on phantoms as well as advanced harmonized quantitative MRI protocols for studying the brain of healthy human participants and wild type mice. Under FAIR principles, a web-based e-infrastructure to store and share data across sites was also implemented (WP4). Finally, the RIN translated all these efforts into a large-scale multimodal data collection in patients and animal models with dementia (i.e., case study). The RIN-Neuroimaging Network can maximize the impact of public investments in research and clinical practice acquiring data across institutes and pathologies with high-quality and highly-consistent acquisition protocols, optimizing the analysis pipeline and data sharing procedures.
| S-EPMC9047871 | biostudies-literature
Project description:With improvements to both scan quality and facial recognition software, there is an increased risk of participants being identified by a 3D render of their structural neuroimaging scans, even when all other personal information has been removed. To prevent this, facial features should be removed before data are shared or openly released, but while there are several publicly available software algorithms to do this, there has been no comprehensive review of their accuracy within the general population. To address this, we tested multiple algorithms on 300 scans from three neuroscience research projects, funded in part by the Ontario Brain Institute, to cover a wide range of ages (3-85 years) and multiple patient cohorts. While skull stripping is more thorough at removing identifiable features, we focused mainly on defacing software, as skull stripping also removes potentially useful information, which may be required for future analyses. We tested six publicly available algorithms (afni_refacer, deepdefacer, mri_deface, mridefacer, pydeface, quickshear), with one skull stripper (FreeSurfer) included for comparison. Accuracy was measured through a pass/fail system with two criteria; one, that all facial features had been removed and two, that no brain tissue was removed in the process. A subset of defaced scans were also run through several preprocessing pipelines to ensure that none of the algorithms would alter the resulting outputs. We found that the success rates varied strongly between defacers, with afni_refacer (89%) and pydeface (83%) having the highest rates, overall. In both cases, the primary source of failure came from a single dataset that the defacer appeared to struggle with - the youngest cohort (3-20 years) for afni_refacer and the oldest (44-85 years) for pydeface, demonstrating that defacer performance not only depends on the data provided, but that this effect varies between algorithms. While there were some very minor differences between the preprocessing results for defaced and original scans, none of these were significant and were within the range of variation between using different NIfTI converters, or using raw DICOM files.
| S-EPMC7943842 | biostudies-literature