Method for early prediction of neurodegenerative decline

A method for training a classifier to learn correlated features, using a plurality of previously-acquired medical images of one or more areas of the brain and previously-acquired clinical and/or biological data, the method comprising:

• extracting one or more texture features from the previously-acquired medical images to obtain one or more extracted texture features,
• correlating said one or more extracted texture features with at least one of the previously-acquired clinical and/or biological data to obtain correlated features, and
• training the classifier to learn said correlated features.

Computer program product for predicting risks of neurodegenerative decline based on at least one medical brain image and at least one clinical and/or biological data of said patient, and using at least a classifier, trained beforehand to learn texture features extracted from a plurality of previously-acquired medical images of one or more areas of the brain and correlated with previously-acquired clinical and/or biological data, the computer program product comprising a support and stored on this support instructions that can be read by a processor, these instructions being configured for:

• extracting one or more texture features from said at least one medical brain image to obtain one or more extracted texture features,
• correlating said one or more extracted texture features with said at least one clinical and/or biological data to obtain one or more correlated features, and
• based at least on said one or more correlated features, operating the trained classifier on the at least one medical brain image to generate a score representative of the risks of neurodegenerative decline of the patient.