Method and system for generating multiparametric nosological images

A method of generating multi-parametric nosological images, comprising the steps of:

• a) obtaining a stack of medical images;
• b) processing the stack of medical images, through the steps of:
• b1) subjecting each image to a noise filtering;
• b2) registering each image to a reference space;
• b3) eliminating from at least some of the images areas corresponding to peripheral tissues of a region of interest;
• b4) correcting inhomogeneities in at least some of the images;
• c) reducing the stack of images to a new representation space of smaller dimension, which maintains or increases discriminative properties of original data in the stack of images;
• d) classifying in an unsupervised manner units of the stack of images obtained in step c) according to a pre-established number of classes and without biological interpretation of the units of the stack of images obtained in step c);
• e) automatically assigning multi-parametric profiles to the classes obtained in step d), through the steps of:
• e1) registering in a non-linear manner a reference template of a region of interest to the stack of images;
• e2) correcting probability maps of healthy tissues of the reference template in pathological areas;
• e3) identifying which classes obtained in step d) represent non-pathological tissues by using the probability maps of healthy tissues corrected in step e2);
• e4) eliminating peripheral and minimal classes from the set of classes obtained in step d) to produce a remaining set of classes;
• and
• e5) merging in the pathological areas the remaining set of classes from step d) through agglomerative hierarchical classification using distances between probability density functions;
• and
• f) generating a nosological image through the classes obtained in steps e3) and e5).