RESUMO
BACKGROUND AND PURPOSE: Hereditary diffuse leukoencephalopathy with spheroids (HDLS) and multiple sclerosis (MS) are demyelinating and neurodegenerative disorders that can be hard to distinguish clinically and radiologically. HDLS is a rare disorder compared to MS, which has led to occurrent misdiagnosis of HDLS as MS. That is problematic since their prognosis and treatment differ. Both disorders are investigated by MRI, which could help to identify patients with high probability of having HDLS, which could guide targeted genetic testing to confirm the HDLS diagnosis. METHODS: Here, we present a machine learning method based on quantitative MRI that can achieve a robust classification of HDLS versus MS. Four HDLS and 14 age-matched MS patients underwent a quantitative brain MRI protocol (synthetic MRI) at 3 Tesla (T) (scan time <7 minutes). We also performed a repeatability analysis of the predicting features to assess their generalizability by scanning a healthy control with five scan-rescans at 3T and 1.5T. RESULTS: Our predicting features were measured with an average confidence interval of 1.7% (P = .01), at 3T and 2.3% (P = .01) at 1.5T. The model gave a 100% correct classification of the cross-validation data when using 5-11 predicting features. When the maximum measurement noise was inserted in the model, the true positive rate of HDLS was 97.2%, while the true positive rate of MS was 99.6%. CONCLUSIONS: This study suggests that computer-assistance in combination with quantitative MRI may be helpful in aiding the challenging differential diagnosis of HDLS versus MS.
Assuntos
Encéfalo/diagnóstico por imagem , Leucoencefalopatias/diagnóstico por imagem , Aprendizado de Máquina , Esclerose Múltipla/diagnóstico por imagem , Adulto , Idoso , Diagnóstico Diferencial , Feminino , Humanos , Imageamento por Ressonância Magnética , Masculino , Pessoa de Meia-IdadeRESUMO
Segmentation of axon and myelin from microscopy images of the nervous system provides useful quantitative information about the tissue microstructure, such as axon density and myelin thickness. This could be used for instance to document cell morphometry across species, or to validate novel non-invasive quantitative magnetic resonance imaging techniques. Most currently-available segmentation algorithms are based on standard image processing and usually require multiple processing steps and/or parameter tuning by the user to adapt to different modalities. Moreover, only a few methods are publicly available. We introduce AxonDeepSeg, an open-source software that performs axon and myelin segmentation of microscopic images using deep learning. AxonDeepSeg features: (i) a convolutional neural network architecture; (ii) an easy training procedure to generate new models based on manually-labelled data and (iii) two ready-to-use models trained from scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Results show high pixel-wise accuracy across various species: 85% on rat SEM, 81% on human SEM, 95% on mice TEM and 84% on macaque TEM. Segmentation of a full rat spinal cord slice is computed and morphological metrics are extracted and compared against the literature. AxonDeepSeg is freely available at https://github.com/neuropoly/axondeepseg .