Publications

@article{gonzalez-castro_reliability_2017,

title = {Reliability of an automatic classifier for brain enlarged perivascular spaces burden and comparison with human performance},

author = {Víctor González-Castro and María del Carmen Valdés-Hernández and Francesca M Chappell and Paul A Armitage and Stephen Makin and Joanna M Wardlaw},

url = {https://portlandpress.com/clinsci/article/131/13/1465/71656/Reliability-of-an-automatic-classifier-for-brain},

year  = {2017},

date = {2017-01-01},

journal = {Clinical Science},

volume = {131},

number = {13},

pages = {1465–1481},

abstract = {Enlarged perivascular spaces (PVS) in the brain are associated with small vessel disease, poor cognition, and hypertension. This study proposes a fully automated method using a support vector machine (SVM) to classify PVS burden in the basal ganglia (BG) as low or high from T2-weighted MRI images. Three feature extraction techniques were evaluated, with the bag of visual words (BoW) approach achieving the highest accuracy (81.16%). The classifier's performance was comparable to that of trained human observers, and its predictions were clinically meaningful, as indicated by high AUC values (0.90–0.93). These findings suggest that automated PVS burden assessment could serve as a valuable clinical tool.},

note = {Publisher: Portland Press Ltd.},

keywords = {Bag of Visual Words, brain MRI, Discrete Wavelet Transform, Local Binary Patterns, machine learning, perivascular spaces, small vessel disease, support vector machine},

pubstate = {published},

tppubtype = {article}

}

@article{gonzalez-castro_automatic_2016,

title = {Automatic rating of perivascular spaces in brain MRI using bag of visual words},

author = {Víctor González-Castro and María del Carmen Valdés-Hernández and Paul A Armitage and Joanna M Wardlaw},

url = {https://link.springer.com/chapter/10.1007/978-3-319-41501-7_72},

year  = {2016},

date = {2016-01-01},

urldate = {2016-01-01},

journal = {Image Analysis and Recognition: 13th International Conference, ICIAR 2016, in Memory of Mohamed Kamel, Póvoa de Varzim, Portugal, July 13-15, 2016, Proceedings 13},

pages = {642–649},

abstract = {This paper presents a fully automatic method for assessing perivascular space (PVS) burden in the basal ganglia using structural MRI. A Support Vector Machine classifier, combined with a Bag of Visual Words (BoW) model, describes the region using two local descriptor approaches: SIFT and textons. The method achieves an accuracy of 82.34% with SIFT and 79.61% with textons, aiding in the study of neurological conditions linked to enlarged PVS.},

note = {Publisher: Springer International Publishing},

keywords = {machine learning, MRI, neurological disorders, perivascular spaces},

pubstate = {published},

tppubtype = {article}

}

@article{ballerini_application_2016,

title = {Application of the ordered logit model to optimising Frangi filter parameters for segmentation of perivascular spaces},

author = {Lucía Ballerini and Ruggiero Lovreglio and María del Carmen Valdés-Hernández and Víctor González-Castro and Susana Muñoz-Maniega and Enrico Pellegrini and Mark E Bastin and Ian J Deary and Joanna M Wardlaw},

url = {https://www.sciencedirect.com/science/article/pii/S1877050916311899},

year  = {2016},

date = {2016-01-01},

journal = {Procedia Computer Science},

volume = {90},

pages = {61–67},

abstract = {Segmenting perivascular spaces (PVS) in brain MRI is crucial for studying the brain's lymphatic system and its link to neurological diseases. The Frangi filter is a useful tool for this task, but its parameters must be optimized for different scanner settings. This study employs an ordered logit model to refine these parameters based on neuroradiological PVS ratings. The resulting PVS volume strongly correlates with expert assessments (Spearman’s ρ=0.75, p < 0.001), indicating that this approach is a promising alternative to conventional optimization methods.},

note = {Publisher: Elsevier},

keywords = {brain MRI, frangi filter, medical imaging, neuroimaging, perivascular spaces},

pubstate = {published},

tppubtype = {article}

}

@article{gonzalez-castro_texture-based_2016,

title = {Texture-based classification for the automatic rating of the perivascular spaces in brain MRI},

author = {Víctor González-Castro and María del Carmen Valdés-Hernández and Paul A Armitage and Joanna M Wardlaw},

url = {https://www.sciencedirect.com/science/article/pii/S1877050916311802},

year  = {2016},

date = {2016-01-01},

journal = {Procedia Computer Science},

volume = {90},

pages = {9–14},

abstract = {This paper investigates the classification of enlarged perivascular spaces (PVS) in the basal ganglia (BG) using texture features extracted from structural brain MRI. The texture is described through first-order statistics, co-occurrence matrix features, discrete wavelet transform coefficients (WSF and WCF), and local binary patterns (LBP). The texture features are classified using a Support Vector Machine (SVM). Experimental results show that WCF provides an accuracy of 80.03% in classifying the density of enlarged PVS.},

note = {Publisher: Elsevier},

keywords = {Discrete Wavelet Transform, local binary pattern, perivascular spaces, support vector machine, texture descriptors},

pubstate = {published},

tppubtype = {article}

}