Challenges and Opportunities in dMRI Data Harmonization

Validation of deep learning techniques for quality augmentation in diffusion MRI for clinical studies

The objective of this study is to evaluate the efficacy of deep learning (DL) techniques in improving the quality of diffusion MRI (dMRI) data in clinical applications. The study aims to determine whether the use of artificial intelligence (AI) methods in medical images may result in the loss of critical clinical information and/or the appearance of false information. To assess this, the focus was on the angular resolution of dMRI and a clinical trial was conducted on migraine, specifically between episodic and chronic migraine patients. The number of gradient directions had an impact on white matter analysis results, with statistically significant differences between groups being drastically reduced when using 21 gradient directions instead of the original 61. Fourteen teams from different institutions were tasked to use DL to enhance three diffusion metrics (FA, AD and MD) calculated from data acquired with 21 gradient directions and a b-value of 1000 s/mm2. The goal was to produce results that were comparable to those calculated from 61 gradient directions. The results were evaluated using both standard image quality metrics and Tract-Based Spatial Statistics (TBSS) to compare episodic and chronic migraine patients. The study results suggest that while most DL techniques improved the ability to detect statistical differences between groups, they also led to an increase in false positive. The results showed that there was a constant growth rate of false positives linearly proportional to the new true positives, which highlights the risk of generalization of AI-based tasks when assessing diverse clinical cohorts and training using data from a single group. The methods also showed divergent performance when replicating the original distribution of the data and some exhibited significant bias. In conclusion, extreme caution should be exercised when using AI methods for harmonization or synthesis in clinical studies when processing heterogeneous data in clinical studies, as important information may be altered, even when global metrics such as structural similarity or peak signal-to-noise ratio appear to suggest otherwise.

Ten quick tips for computational analysis of medical images

Medical imaging is a great asset for modern medicine, since it allows physicians to spatially interrogate a disease site, resulting in precise intervention for diagnosis and treatment, and to observe particular aspect of patients' conditions that otherwise would not be noticeable. Computational analysis of medical images, moreover, can allow the discovery of disease patterns and correlations among cohorts of patients with the same disease, thus suggesting common causes or providing useful information for better therapies and cures. Machine learning and deep learning applied to medical images, in particular, have produced new, unprecedented results that can pave the way to advanced frontiers of medical discoveries. While computational analysis of medical images has become easier, however, the possibility to make mistakes or generate inflated or misleading results has become easier, too, hindering reproducibility and deployment. In this article, we provide ten quick tips to perform computational analysis of medical images avoiding common mistakes and pitfalls that we noticed in multiple studies in the past. We believe our ten guidelines, if taken into practice, can help the computational-medical imaging community to perform better scientific research that eventually can have a positive impact on the lives of patients worldwide.

Harmonization of Multi-Center Diffusion Tensor Tractography in Neonates with Congenital Heart Disease: Optimizing Post-Processing and Application of ComBat

Advanced brain imaging of neonatal macrostructure and microstructure, which has prognosticating importance, is more frequently being incorporated into multi-center trials of neonatal neuroprotection. Multicenter neuroimaging studies, designed to overcome small sample sized clinical cohorts, are essential but lead to increased technical variability. Few harmonization techniques have been developed for neonatal brain microstructural (diffusion tensor) analysis. The work presented here aims to remedy two common problems that exist with the current state of the art approaches: 1) variance in scanner and protocol in data collection can limit the researcher's ability to harmonize data acquired under different conditions or using different clinical populations. 2) The general lack of objective guidelines for dealing with anatomically abnormal anatomy and pathology. Often, subjects are excluded due to subjective criteria, or due to pathology that could be informative to the final analysis, leading to the loss of reproducibility and statistical power. This proves to be a barrier in the analysis of large multi-center studies and is a particularly salient problem given the relative scarcity of neonatal imaging data. We provide an objective, data-driven, and semi-automated neonatal processing pipeline designed to harmonize compartmentalized variant data acquired under different parameters. This is done by first implementing a search space reduction step of extracting the along-tract diffusivity values along each tract of interest, rather than performing whole-brain harmonization. This is followed by a data-driven outlier detection step, with the purpose of removing unwanted noise and outliers from the final harmonization. We then use an empirical Bayes harmonization algorithm performed at the along-tract level, with the output being a lower dimensional space but still spatially informative. After applying our pipeline to this large multi-site dataset of neonates and infants with congenital heart disease (n= 398 subjects recruited across 4 centers, with a total of n=763 MRI pre-operative/post-operative time points), we show that infants with single ventricle cardiac physiology demonstrate greater white matter microstructural alterations compared to infants with bi-ventricular heart disease, supporting what has previously been shown in literature. Our method is an open-source pipeline for delineating white matter tracts in subject space but provides the necessary modular components for performing atlas space analysis. As such, we validate and introduce Diffusion Imaging of Neonates by Group Organization (DINGO), a high-level, semi-automated framework that can facilitate harmonization of subject-space tractography generated from diffusion tensor imaging acquired across varying scanners, institutions, and clinical populations. Datasets acquired using varying protocols or cohorts are compartmentalized into subsets, where a cohort-specific template is generated, allowing for the propagation of the tractography mask set with higher spatial specificity. Taken together, this pipeline can reduce multi-scanner technical variability which can confound important biological variability in relation to neonatal brain microstructure.

The ENIGMA Stroke Recovery Working Group: Big data neuroimaging to study brain-behavior relationships after stroke

The goal of the Enhancing Neuroimaging Genetics through Meta-Analysis (ENIGMA) Stroke Recovery working group is to understand brain and behavior relationships using well-powered meta- and mega-analytic approaches. ENIGMA Stroke Recovery has data from over 2,100 stroke patients collected across 39 research studies and 10 countries around the world, comprising the largest multisite retrospective stroke data collaboration to date. This article outlines the efforts taken by the ENIGMA Stroke Recovery working group to develop neuroinformatics protocols and methods to manage multisite stroke brain magnetic resonance imaging, behavioral and demographics data. Specifically, the processes for scalable data intake and preprocessing, multisite data harmonization, and large-scale stroke lesion analysis are described, and challenges unique to this type of big data collaboration in stroke research are discussed. Finally, future directions and limitations, as well as recommendations for improved data harmonization through prospective data collection and data management, are provided.

Pre-statistical harmonization of behavrioal instruments across eight surveys and trials

BACKGROUND

Data harmonization is a powerful method to equilibrate items in measures that evaluate the same underlying construct. There are multiple measures to evaluate dementia related behavioral symptoms. Pre-statistical harmonization of behavioral instruments in dementia research is the first step to develop a statistical crosswalk between measures. Studies that conduct pre-statistical harmonization of behavioral instruments rarely document their methods in a structured, reproducible manner. This is a crucial step which entails careful review, documentation and scrutiny of source data to ensure sufficient comparability between items prior to data pooling. Here, we document the pre-statistical harmonization of items measuring behavioral and psychological symptoms among people with dementia. We provide a box of recommended procedure for future studies.

METHODS

We identified behavioral instruments that are used in clinical practice, a national survey, and randomized trials of dementia care interventions. We rigorously reviewed question content and scoring procedures to establish sufficient comparability across items as well as item quality prior to data pooling. Additionally, we standardized coding to Stata-readable format, which allowed us to automate approaches to identify potential cross-study differences in items and low-quality items. To ensure reasonable model fit for statistical co-calibration, we estimated two-parameter logistic Item Response Theory models within each of the eight studies.

RESULTS

We identified 59 items from 11 behavioral instruments across the eight datasets. We found considerable cross-study heterogeneity in administration and coding procedures for items that measure the same attribute. Discrepancies existed in terms of directionality and quantification of behavioral symptoms for even seemingly comparable items. We resolved item response heterogeneity, missingness and skewness, conditional dependency prior to estimation of item response theory models for statistical co-calibration. We used several rigorous data transformation procedures to address these issues, including re-coding and truncation.

CONCLUSIONS

This study highlights the importance of each aspect involved in the pre-statistical harmonization process of behavioral instruments. We provide guidelines and recommendations for how future research may detect and account for similar issues in pooling behavioral and related instruments.

Combining transcranial magnetic stimulation with functional magnetic resonance imaging for probing and modulating neural circuits relevant to affective disorders

Combining transcranial magnetic stimulation (TMS) with functional magnetic resonance imaging offers an unprecedented tool for studying how brain networks interact in vivo and how repetitive trains of TMS modulate those networks among patients diagnosed with affective disorders. TMS compliments neuroimaging by allowing the interrogation of causal control among brain circuits. Together with TMS, neuroimaging can provide valuable insight into the mechanisms underlying treatment effects and downstream circuit communication. Here we provide a background of the method, review relevant study designs, consider methodological and equipment options, and provide statistical recommendations. We conclude by describing emerging approaches that will extend these tools into exciting new applications. This article is categorized under: Psychology > Emotion and Motivation Psychology > Theory and Methods Neuroscience > Clinical Neuroscience.

Challenges and opportunities for neuroimaging in young patients with traumatic brain injury: a coordinated effort towards advancing discovery from the ENIGMA pediatric moderate/severe TBI group

Traumatic brain injury (TBI) is a major cause of death and disability in children in both developed and developing nations. Children and adolescents suffer from TBI at a higher rate than the general population, and specific developmental issues require a unique context since findings from adult research do not necessarily directly translate to children. Findings in pediatric cohorts tend to lag behind those in adult samples. This may be due, in part, both to the smaller number of investigators engaged in research with this population and may also be related to changes in safety laws and clinical practice that have altered length of hospital stays, treatment, and access to this population. The ENIGMA (Enhancing NeuroImaging Genetics through Meta-Analysis) Pediatric Moderate/Severe TBI (msTBI) group aims to advance research in this area through global collaborative meta-analysis of neuroimaging data. In this paper, we discuss important challenges in pediatric TBI research and opportunities that we believe the ENIGMA Pediatric msTBI group can provide to address them. With the paucity of research studies examining neuroimaging biomarkers in pediatric patients with TBI and the challenges of recruiting large numbers of participants, collaborating to improve statistical power and to address technical challenges like lesions will significantly advance the field. We conclude with recommendations for future research in this field of study.

Harmonization of diffusion MRI data sets with adaptive dictionary learning

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.

Harmonization of Brain Diffusion MRI: Concepts and Methods

MRI diffusion data suffers from significant inter- and intra-site variability, which hinders multi-site and/or longitudinal diffusion studies. This variability may arise from a range of factors, such as hardware, reconstruction algorithms and acquisition settings. To allow a reliable comparison and joint analysis of diffusion data across sites and over time, there is a clear need for robust data harmonization methods. This review article provides a comprehensive overview of diffusion data harmonization concepts and methods, and their limitations. Overall, the methods for the harmonization of multi-site diffusion images can be categorized in two main groups: diffusion parametric map harmonization (DPMH) and diffusion weighted image harmonization (DWIH). Whereas DPMH harmonizes the diffusion parametric maps (e.g., FA, MD, and MK), DWIH harmonizes the diffusion-weighted images. Defining a gold standard harmonization technique for dMRI data is still an ongoing challenge. Nevertheless, in this paper we provide two classification tools, namely a feature table and a flowchart, which aim to guide the readers in selecting an appropriate harmonization method for their study.