20 Jul 21. Featured Paper

Link to paper on Springer Nature

Authors

Jose Bernal, William Xu, Maria d. C. Valdés-Hernández, Javier Escudero, Angela C. C. Jochems, Una Clancy, Fergus N. Doubal, Michael S. Stringer, Michael J. Thrippleton, Rhian M. Touyz, Joanna M. Wardlaw

Abstract

Current evidence points towards perivascular spaces playing a key role in cerebral haemodynamics & waste clearance.

Hence, their precise quantification may become a powerful tool for assessing brain health & further establishing their relationship with neurological diseases.

Large strides have been made towards developing automatic tools to computationally assess the burden of perivascular spaces in MRI in recent years.

However, their applicability depends to a large extent on the quality of the images.

In this paper, we propose a pipeline to improve perivascular space quantification by means of radiomics-based image quality control & selective motion artefacts reduction.

We demonstrate our method on a sample of patients with mild stroke (n = 60) with different extents of small vessel disease features & image quality.

We show our proposal can differentiate high- & low-quality scans (AUROC = 0.98) & reduce imaging artefacts, which leads to greater correlations between visual & computational measurements, especially in the centrum semiovale (polyserial correlation: 0.86 [95% CI 0.85, 0.88] & 0.17 [95% CI 0.14, 0.21] with & without our proposal, respectively).

Our preliminary results demonstrate the potential of our proposal for retaining clinically relevant information while reducing imaging artefacts.

Keywords

• Brain magnetic resonance imaging
• Cerebral small vessel disease
• Image enhancement
• Imaging artefact reduction
• Perivascular spaces