Integrated MRI Analysis and Quality Control Pipeline for Memory Clinic Use of UK Biobank Imaging
Poster No:
216
Submission Type:
Abstract Submission
Authors:
Grace Gillis1,2, Gaurav Bhalerao1, Jasmine Blane1,2, Pieter Pretorius1,3, Lola Martos1,2, Vanessa Raymont1,2, Clare Mackay1, Ludovica Griffanti1,2
Institutions:
1Department of Psychiatry, University of Oxford, Oxford, Oxfordshire, 2Oxford Health NHS Foundation Trust, Oxford, United Kingdom, 3Oxford University Hospitals NHS Foundation Trust, Oxford, United Kingdom
First Author:
Grace Gillis
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Co-Author(s):
Jasmine Blane
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Pieter Pretorius
Department of Psychiatry, University of Oxford|Oxford University Hospitals NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Department of Psychiatry, University of Oxford|Oxford University Hospitals NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Lola Martos
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Vanessa Raymont
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Ludovica Griffanti
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Department of Psychiatry, University of Oxford|Oxford Health NHS Foundation Trust
Oxford, Oxfordshire|Oxford, United Kingdom
Introduction:
Sophisticated imaging protocols and analysis techniques have been developed in research contexts to extract metrics known as imaging-derived phenotypes (IDPs). However, it remains unclear whether these methods can also yield accurate and meaningful measures when applied in a clinical setting. In a clinical context, it is also essential to perform quality control (QC) in parallel with any analyses to inform the interpretation of the generated metrics. Therefore, in this study we aimed to adapt the UK Biobank (UKB) MRI analysis pipeline, assess its performance in a memory clinic setting (the Oxford Brain Health Clinic), and provide an integrated analysis and QC pipeline for use in the memory clinic.
Methods:
As part of their memory clinic assessment at the Oxford Brain Health Clinic (O'Donoghue et al., 2023), 213 patients were scanned using an adapted version of the UKB protocol [T1-weighted, T2-FLAIR, susceptibility-weighted (swMRI), quantitative susceptibility mapping (QSM), diffusion-weighted (dMRI), arterial spin labelling (ASL), and resting-state functional MRI (rfMRI)] (Miller et al., 2016; Griffanti et al., 2022). As previously described, the UKB processing pipeline was adapted to include lesion-masking of the SIENAX grey matter segmentations and CSF-masking of the FIRST hippocampal segmentations (Griffanti et al., 2022). Downstream pipeline components reliant on these corrected segmentations were also adapted, and white matter hyperintensities (WMHs) were further classified into periventricular and deep WMHs, in line with neuroradiologist-reported metrics (Figure 1). Quality control (QC) was performed on the raw scans and pipeline outputs to assess the quality of the acquired data and explore whether additional pipeline modifications may be necessary for this clinical application. Although supplemented by visual QC where necessary (SWI, QSM, and ASL), automated tools were used where possible for the first-pass QC: MRIQC (Esteban et al., 2017) for T1-weighted and T2-FLAIR scans, QUAD (Bastiani et al., 2019) for dMRI, and MRIQC and DSE decomposition (Afyouni and Nichols, 2018) for rfMRI. The core outputs from all flagged scans were visually inspected. We investigated the associations of IDPs with diagnoses and cognitive scores (ACE-III) in this unselected memory clinic population.
·Figure 1: Simplified overview of the UK Biobank image analysis pipeline with adaptations for use in the Oxford Brain Health Clinic.
Results:
QC results are summarised in Table 1. MRIQC was capable of flagging T1-weighted and T2-FLAIR scans for further inspection, but the adapted pipeline still generated mostly high- or medium-quality outputs in these scans (96.8%, 93.5%, and 94.6% for grey matter, hippocampal, and WMH segmentations, respectively). QUAD was able to flag lower quality dMRI scans, but visual inspection revealed that all of the flagged scans had high (66.7%) or medium-quality (33.3%) tractography results. DSE decomposition and MRIQC together could identify challenging rfMRI scans. Over half of the flagged scans still had significant structured noise present in their processed data, highlighting the need for further optimisation of the rfMRI pipeline for this memory clinic use. Compared to the other T2-FLAIR IDPs, periventricular WMH volume associated most strongly with cognition and diagnoses, supporting the use of this additional metric in the memory clinic setting.
·Table 1: Results of first-pass and detailed visual quality control for T1-weighted, T2-FLAIR, dMRI, and rfMRI scans.
Conclusions:
We adapted research-quality MRI acquisition and processing, aligned with the UK Biobank, into the memory clinic setting at the Oxford Brain Health Clinic. We integrated the analysis and quality control steps into a processing pipeline for clinical use and have explored its value to extend research findings into an unselected clinical population.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Lifespan Development:
Aging
Modeling and Analysis Methods:
Diffusion MRI Modeling and Analysis
Segmentation and Parcellation
Neuroinformatics and Data Sharing:
Workflows 2
Keywords:
Aging
Cerebrovascular Disease
Cognition
Data analysis
Degenerative Disease
DISORDERS
STRUCTURAL MRI
Workflows
Other - Clinical Applications; Quality Control
1|2Indicates the priority used for review
Provide references using author date format
Bastiani, M. et al. (2019) ‘Automated quality control for within and between studies diffusion MRI data using a non-parametric framework for movement and distortion correction’, Neuroimage, 184, pp. 801–812.
Esteban, O. et al. (2017) ‘MRIQC: Advancing the automatic prediction of image quality in MRI from unseen sites’, PLOS ONE, 12(9), p. e0184661.
Griffanti, L. et al. (2022) ‘Adapting UK Biobank imaging for use in a routine memory clinic setting: The Oxford Brain Health Clinic’, NeuroImage: Clinical, 36, p. 103273.
Miller, K.L. et al. (2016) ‘Multimodal population brain imaging in the UK Biobank prospective epidemiological study’, Nature Neuroscience, 19(11), pp. 1523–1536.
O’Donoghue, M.C. et al. (2023) ‘Oxford brain health clinic: protocol and research database’, BMJ Open, 13(8), p. e067808.