Assessment of perivascular space filtering methods using a three-dimensional computational model: source code

Publication available; If you use our computational model in your research, please cite the following paper: Bernal J, Valdés-Hernández MD, Escudero J, Duarte R, Ballerini L, Bastin ME, Deary IJ, Thrippleton MJ, Touyz RM, Wardlaw JM. Assessment of perivascular space filtering methods using a three-dimensional computational model. Magnetic resonance imaging. 2022 Nov 1;93:33-51. Introduction ------------ Perivascular spaces (PVS) are fluid-filled tubular passageways surrounding cerebral microvessels, thought to facilitate waste clearance and fluid exchange. PVS can become visible on structural MRI and, while a few may be normal, increased number or size is linked to ageing, hypertension, blood-brain barrier dysfunction, and other small vessel disease (SVD) markers. Their accurate quantification is important for understanding SVD and broader neurological health. Materials and methods ------------ Computational model for assessing PVS quantification. We generated maps with PVS-like structures in different orientations and locations throughout the brain. We used a high-resolution model of the head and the resulting “PVS” maps to synthesise T2-weighted like images. We then sampled these high-resolution digital reference objects (DRO) and incorporated motion and Rician noise to generate the “acquired” T2-weighted images. We finally segmented the “PVS” on these DRO and compared the results against the ground truth. The closer the results to the ground truth, the more resilient the “vesselness” filtering method against distortion. Result ------------ Our findings were three-fold. First, as long as voxels are isotropic, RORPO outperforms the other two filters, regardless of imaging quality. Unlike the Frangi and Jerman filters, RORPO's performance does not deteriorate as PVS volume increases. Second, the performance of all “vesselness” filters is heavily influenced by imaging quality, with sampling and motion artefacts being the most damaging for these types of analyses. Third, none of the filters can distinguish PVS from other hyperintense structures (e.g. white matter hyperintensities, stroke lesions, or lacunes) effectively, the area under precision-recall curve dropped substantially (Frangi: from 94.21 [IQR 91.60, 96.16] to 43.76 [IQR 25.19, 63.38]; Jerman: from 94.51 [IQR 91.90, 95.37] to 58.00 [IQR 35.68, 64.87]; RORPO: from 98.72 [IQR 95.37, 98.96] to 71.87 [IQR 57.21, 76.63] without and with other hyperintense structures, respectively). Conclusions ------------ Our work reveals appropriate processing of MRI signals is necessary to maximise PVS measurement reliability. While filters to enhance and facilitate PVS detection can correctly achieve these goals, they are sensitive to imaging artefacts, such as ringing and motion, and ineffective at distinguishing between PVS from other lesions with similar contrast, shapes and overlapping size. These issues ultimately highlight the importance of masking out other neuroradiological features and of prospective or retrospective image enhancement for better PVS quantification. Computational model from the article: Bernal, Jose, Maria DC Valdés-Hernández, Javier Escudero, Roberto Duarte, Lucia Ballerini, Mark E. Bastin, Ian J. Deary, Michael J. Thrippleton, Rhian M. Touyz, and Joanna M. Wardlaw. "Assessment of perivascular space filtering methods using a three-dimensional computational model." Magnetic resonance imaging 93 (2022): 33-51. Acknowledgements ------------ This work was supported by: MRC Doctoral Training Programme in Precision Medicine (JB - Award Reference No. 2096671); The Galen and Hilary Weston Foundation under the Novel Biomarkers 2019 scheme (ref UB190097) administered by the Weston Brain Institute; the UK Dementia Research Institute which receives its funding from DRI Ltd., funded by the UK MRC, Alzheimer's Society and Alzheimer's Research UK; the Fondation Leducq Network for the Study of Perivascular Spaces in Small Vessel Disease (16 CVD 05); Stroke Association ‘Small Vessel Disease-Spotlight on Symptoms (SVD-SOS)’ (SAPG 19\100068); The Row Fogo Charitable Trust Centre for Research into Ageing and the Brain (MVH) (BRO-D.FID3668413); British Heart Foundation Edinburgh Centre for Research Excellence (RE/18/5/34216); NHS Lothian Research and Development Office (MJT).
The LBC1936 study was funded by Age UK and the UK Medical Research Council (http://www.disconnectedmind.ed.ac.uk/) (including the Sidney De Haan Award for Vascular Dementia). LBC1936 MRI brain imaging was supported by Medical Research Council (MRC) grants G0701120, G1001245, MR/M013111/1 and MR/R024065/1. Funds were also received from The University of Edinburgh Centre for Cognitive Ageing and Cognitive Epidemiology, part of the cross-council Lifelong Health and Wellbeing Initiative (MR/K026992/1), and the Biotechnology and Biological Sciences Research Council (BBSRC).
We thank the LBC1936 study participants, their families and radiographers at Edinburgh Imaging Facilities. We thank the LBC1936 study team members who recruited the participants of the study.