Method and system for automated central vein sign assessment

A method for automatically detecting, in a magnetic resonance image, a white matter lesion exhibiting a central vein sign, the method comprising steps of:

• acquiring a set including one or several 3D MR images of a brain lesion;
• using a system for automated assessment of the central vein sign, the system including a processing unit and a memory, and the system:
• using a first and a second subset of the set as a respective input to a first and a second machine learning algorithm, the first machine learning algorithm and the second machine learning algorithm being two different machine learning algorithms;
• configuring the first machine learning algorithm for performing a classification of the first subset into a first class or a second class, the first class including central vein sign +lesions and central vein sign −lesions, and the second class including central vein sign e lesions, the first machine learning algorithm being further configured for outputting, for the first subset received as an input and each class, a probability that the subset belongs to the class;
• configuring the second machine learning algorithm for performing a classification of the second subset into a third class or a fourth class, the third class including central vein sign +lesions and the fourth class including central vein sign −lesions, the second machine learning algorithm being further configured for outputting for the second subset received as an input and each class, a probability that the second subset belongs to the class;
• for each set, using probability results obtained from the first and second machine learning algorithms as inputs to a final classifier configured for performing a final classification of the set into the second, third, or fourth class, and for each of the classes, configuring the final classifier for outputting a final probability that the set belongs to the class;
• and
• providing, through an interface, a class, among the second, third and fourth classes, being characterized by a highest probability value.