Corpus ID: 204009031

Machine Learning with Multi-Site Imaging Data: An Empirical Study on the Impact of Scanner Effects

@article{Glocker2019MachineLW,
  title={Machine Learning with Multi-Site Imaging Data: An Empirical Study on the Impact of Scanner Effects},
  author={Ben Glocker and Robert Robinson and Daniel Coelho de Castro and Qi Dou and Ender Konukoglu},
  journal={ArXiv},
  year={2019},
  volume={abs/1910.04597}
}
This is an empirical study to investigate the impact of scanner effects when using machine learning on multi-site neuroimaging data. We utilize structural T1-weighted brain MRI obtained from two different studies, Cam-CAN and UK Biobank. For the purpose of our investigation, we construct a dataset consisting of brain scans from 592 age- and sex-matched individuals, 296 subjects from each original study. Our results demonstrate that even after careful pre-processing with state-of-the-art… Expand
Deep Learning-Based Unlearning of Dataset Bias for MRI Harmonisation and Confound Removal
TLDR
A deep learning based training scheme, inspired by domain adaptation techniques, which uses an iterative update approach to aim to create scanner-invariant features while simultaneously maintaining performance on the main task of interest, thus reducing the influence of scanner on network predictions. Expand
Removal of Scanner Effects in Covariance Improves Multivariate Pattern Analysis in Neuroimaging Data
TLDR
It is demonstrated that methods for removing scanner effects in mean and variance may not be sufficient for multivariate pattern analysis, and a novel methodology is proposed that harmonizes covariance of multivariate image measurements across scanners and demonstrates its improved performance in data harmonization. Expand
Brain-Age Prediction Using Shallow Machine Learning: Predictive Analytics Competition 2019
TLDR
An ensemble of shallow machine learning methods that combined voxel-based and surface-based morphometric data and the use of genetic algorithms and grid search was developed, showing the potential that shallow methods still have in predicting brain-age. Expand
Robust White Matter Hyperintensity Segmentation On Unseen Domain
TLDR
This work considers the challenging case of Domain Generalization (DG) where a model is trained without any knowledge about the testing distribution, and identifies how two mechanically distinct DG approaches, namely domain adversarial learning and mix-up, have theoretical synergy. Expand
Local brain-age: A U-Net model
TLDR
A U-Net model is trained using brain MRI scans from n=3463 healthy people to produce individualised 3D maps of brain-predicted age and a new voxelwise method is introduced to reduce the age-bias when predicting local brain-age “gaps”. Expand
Benchmarking CNN on 3D Anatomical Brain MRI: Architectures, Data Augmentation and Deep Ensemble Learning
TLDR
It is found that all models provide significantly better predictions with VBM images than quasi-raw data, and it is demonstrated that DenseNet and tiny-DenseNet, a lighter version that is proposed, provide a good compromise in terms of performance in all data regime. Expand
MS-Net: Multi-Site Network for Improving Prostate Segmentation With Heterogeneous MRI Data
TLDR
A novel multi-site network (MS-Net) is proposed for improving prostate segmentation by learning robust representations, leveraging multiple sources of data and develops Domain-Specific Batch Normalization layers in the network backbone to compensate for inter-site heterogeneity of different MRI datasets. Expand
Integrating large-scale neuroimaging research datasets: harmonisation of white matter hyperintensity measurements across Whitehall and UK Biobank datasets
TLDR
This work harmonised measures of WMHs across two studies on healthy ageing and identifies pre-processing strategies that maximise the consistency across datasets and utilise multivariate regression to characterise sample differences contributing to study-level differences in WMH variations. Expand
Contrast Adaptive Tissue Classification by Alternating Segmentation and Synthesis
TLDR
This work describes an approach using alternating segmentation and synthesis steps that adapts the contrast properties of the training data to the input image and demonstrates the efficacy of this approach using brain images from a set of human subjects scanned with two different T1-weighted volumetric protocols. Expand
Complex Grey Matter Structure Segmentation in Brains via Deep Learning: Example of the Claustrum
TLDR
It is found that AI-based claustrum segmentation benefits from multi-view information and requires sample sizes of around 75 MRI scans in the training set, which represents equal or superior segmentation performance compared to human intra-rater reliability. Expand
...
1
2
3
4
...

References

SHOWING 1-10 OF 17 REFERENCES
Harmonization of cortical thickness measurements across scanners and sites
TLDR
It is shown that ComBat removes unwanted sources of scan variability while simultaneously increasing the power and reproducibility of subsequent statistical analyses, and is useful for combining imaging data with the goal of studying life-span trajectories in the brain. Expand
Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank
TLDR
The pipeline is described in detail, following a brief overview of UK Biobank brain imaging and the acquisition protocol and several quantitative investigations carried out as part of the development of both the imaging protocol and the processing pipeline. Expand
Harmonization of multi-site diffusion tensor imaging data
TLDR
It is shown that the DTI measurements are highly site‐specific, highlighting the need of correcting for site effects before performing downstream statistical analyses, and that ComBat, a popular batch‐effect correction tool used in genomics, performs best at modeling and removing the unwanted inter‐site variability in FA and MD maps. Expand
Multimodal population brain imaging in the UK Biobank prospective epidemiological study
TLDR
UK Biobank brain imaging is described and results derived from the first 5,000 participants' data release are presented, which have already yielded a rich range of associations between brain imaging and other measures collected by UK Biobanks. Expand
Statistical Parametric Mapping: The Analysis of Functional Brain Images
In an age where the amount of data collected from brain imaging is increasing constantly, it is of critical importance to analyse those data within an accepted framework to ensure proper integrationExpand
Robust Brain Extraction Across Datasets and Comparison With Publicly Available Methods
TLDR
This paper introduces a robust, learning-based brain extraction system (ROBEX), which combines a discriminative and a generative model to achieve the final result and shows that ROBEX provides significantly improved performance measures for almost every method/dataset combination. Expand
Quantifying Confounding Bias in Neuroimaging Datasets with Causal Inference
TLDR
This work combines 12,207 MRI scans from 15 studies and shows that simple pooling is often ill-advised due to introducing various types of biases in the training data, and proposes to tell causal from confounding factors by quantifying the extent of confounding and causality in a single dataset using causal inference. Expand
New variants of a method of MRI scale standardization
TLDR
New variants of this standardizing method can help to overcome some of the problems with the original method and extraction of quantitative information about healthy organs or about abnormalities can be considerably simplified. Expand
N4ITK: Improved N3 Bias Correction
TLDR
A variant of the popular nonparametric nonuniform intensity normalization (N3) algorithm is proposed for bias field correction with the substitution of a recently developed fast and robust B-spline approximation routine and a modified hierarchical optimization scheme for improved bias field Correction over the original N3 algorithm. Expand
The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) data repository: Structural and functional MRI, MEG, and cognitive data from a cross-sectional adult lifespan sample
TLDR
The Cam-CAN Stage 2 repository contains multi-modal (MRI, MEG, and cognitive-behavioural) data from a large, cross-sectional adult lifespan (18–87 years old) population-based sample, providing a depth of neurocognitive phenotyping that is currently unparalleled, enabling integrative analyses of age-related changes in brain structure, brain function, and cognition. Expand
...
1
2
...