Dataset Information

A k-Space Model of Movement Artefacts: Application to Segmentation Augmentation and Artefact Removal.

ABSTRACT: Patient movement during the acquisition of magnetic resonance images (MRI) can cause unwanted image artefacts. These artefacts may affect the quality of clinical diagnosis and cause errors in automated image analysis. In this work, we present a method for generating realistic motion artefacts from artefact-free magnitude MRI data to be used in deep learning frameworks, increasing training appearance variability and ultimately making machine learning algorithms such as convolutional neural networks (CNNs) more robust to the presence of motion artefacts. By modelling patient movement as a sequence of randomly-generated, 'demeaned', rigid 3D affine transforms, we resample artefact-free volumes and combine these in k-space to generate motion artefact data. We show that by augmenting the training of semantic segmentation CNNs with artefacts, we can train models that generalise better and perform more reliably in the presence of artefact data, with negligible cost to their performance on clean data. We show that the performance of models trained using artefact data on segmentation tasks on real-world test-retest image pairs is more robust. We also demonstrate that our augmentation model can be used to learn to retrospectively remove certain types of motion artefacts from real MRI scans. Finally, we show that measures of uncertainty obtained from motion augmented CNN models reflect the presence of artefacts and can thus provide relevant information to ensure the safe usage of deep learning extracted biomarkers in a clinical pipeline.

SUBMITTER: Shaw R

PROVIDER: S-EPMC7116018 | biostudies-literature | 2020 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A k-Space Model of Movement Artefacts: Application to Segmentation Augmentation and Artefact Removal.

Shaw Richard R Sudre Carole H CH Varsavsky Thomas T Ourselin Sebastien S Cardoso M Jorge MJ

IEEE transactions on medical imaging 20200305 9

Patient movement during the acquisition of magnetic resonance images (MRI) can cause unwanted image artefacts. These artefacts may affect the quality of clinical diagnosis and cause errors in automated image analysis. In this work, we present a method for generating realistic motion artefacts from artefact-free magnitude MRI data to be used in deep learning frameworks, increasing training appearance variability and ultimately making machine learning algorithms such as convolutional neural networ ...[more]

PMID: 32149627

Similar Datasets

Project description:Motion artefacts (MAs) are induced within EEG data collected simultaneously with fMRI when the subject's head rotates relative to the magnetic field. The effects of these artefacts have generally been ameliorated by removing periods of data during which large artefact voltages appear in the EEG traces. However, even when combined with other standard post-processing methods, this strategy does not remove smaller MAs which can dominate the neuronal signals of interest. A number of methods are therefore being developed to characterise the MA by measuring reference signals and then using these in artefact correction. These methods generally assume that the head and EEG cap, plus any attached sensors, form a rigid body which can be characterised by a standard set of six motion parameters. Here we investigate the motion of the head/EEG cap system to provide a better understanding of MAs. We focus on the reference layer artefact subtraction (RLAS) approach, as this allows measurement of a separate reference signal for each electrode that is being used to measure brain activity. Through a series of experiments on phantoms and subjects, we find that movement of the EEG cap relative to the phantom and skin on the forehead is relatively small and that this non-rigid body movement does not appear to cause considerable discrepancy in artefacts between the scalp and reference signals. However, differences in the amplitude of these signals is observed which may be due to differences in geometry of the system from which the reference signals are measured compared with the brain signals. In addition, we find that there is non-rigid body movement of the skull and skin which produces an additional MA component for a head shake, which is not present for a head nod. This results in a large discrepancy in the amplitude and temporal profile of the MA measured on the scalp and reference layer, reducing the efficacy of MA correction based on the reference signals. Together our data suggest that the efficacy of the correction of MA using any reference-based system is likely to differ for different types of head movement with head shake being the hardest to correct. This provides new information to inform the development of hardware and post-processing methods for removing MAs from EEG data acquired simultaneously with fMRI data.

Project description:The identification of resting state networks (RSNs) and the quantification of their functional connectivity in resting-state fMRI (rfMRI) are seriously hindered by the presence of artefacts, many of which overlap spatially or spectrally with RSNs. Moreover, recent developments in fMRI acquisition yield data with higher spatial and temporal resolutions, but may increase artefacts both spatially and/or temporally. Hence the correct identification and removal of non-neural fluctuations is crucial, especially in accelerated acquisitions. In this paper we investigate the effectiveness of three data-driven cleaning procedures, compare standard against higher (spatial and temporal) resolution accelerated fMRI acquisitions, and investigate the combined effect of different acquisitions and different cleanup approaches. We applied single-subject independent component analysis (ICA), followed by automatic component classification with FMRIB's ICA-based X-noiseifier (FIX) to identify artefactual components. We then compared two first-level (within-subject) cleaning approaches for removing those artefacts and motion-related fluctuations from the data. The effectiveness of the cleaning procedures was assessed using time series (amplitude and spectra), network matrix and spatial map analyses. For time series and network analyses we also tested the effect of a second-level cleaning (informed by group-level analysis). Comparing these approaches, the preferable balance between noise removal and signal loss was achieved by regressing out of the data the full space of motion-related fluctuations and only the unique variance of the artefactual ICA components. Using similar analyses, we also investigated the effects of different cleaning approaches on data from different acquisition sequences. With the optimal cleaning procedures, functional connectivity results from accelerated data were statistically comparable or significantly better than the standard (unaccelerated) acquisition, and, crucially, with higher spatial and temporal resolution. Moreover, we were able to perform higher dimensionality ICA decompositions with the accelerated data, which is very valuable for detailed network analyses.

Dataset Information

A k-Space Model of Movement Artefacts: Application to Segmentation Augmentation and Artefact Removal.

Publications

A k-Space Model of Movement Artefacts: Application to Segmentation Augmentation and Artefact Removal.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure