3D regression neural network for the quantification of enlarged perivascular spaces in brain MRI

• Published in: Medical image analysis Vol. 51; pp. 89 - 100
• Main Authors: Dubost, Florian, Adams, Hieab, Bortsova, Gerda, Ikram, M. Arfan, Niessen, Wiro, Vernooij, Meike, de Bruijne, Marleen
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.01.2019; Elsevier BV
• Subjects: Artificial neural networks; Automation; Basal ganglia; Brain; Convolution; Correlation analysis; Correlation coefficient; Correlation coefficients; Deep learning; Dementia; Dementia disorders; Etiology; Ganglia; Health risks; Image detection; Image processing; Magnetic resonance imaging; Neural networks; Neuroimaging; Neurological diseases; Object recognition; Perivascular space; Real numbers; Regression; Reproducibility; Test sets; Training; Vascular diseases; Virchow-Robin space; Weak labels; Deep learning; Regression; Weak labels; Perivascular space; Virchow-Robin space; Dementia

•A regression neural network to quantify enlarged perivascular spaces in brain MRI.•Successful network training based on only the number of lesions in a single slice.•Extensive evaluation on 400 images.•The correlation between automated and visual scores is close to intrarater agreement.•The correla...