An Approach with Neural Network to Detection of MRI Anatomy
- Resource Type
- Authors
- Guido Pasquariello; A. Giancaspro; D. Milella; A. Giannini; A. Blonda; Aristide Carella; Giuseppe Satalino; Franca Dicuonzo; F. Ambrosi; R. De Blasi
- Source
- Scopus-Elsevier
- Subject
- Self-organizing map
Radiological and Ultrasound Technology
Artificial neural network
business.industry
Computer science
ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION
Process (computing)
Pattern recognition
Image segmentation
Object (computer science)
ComputingMethodologies_PATTERNRECOGNITION
Section (archaeology)
Multilayer perceptron
Learning rule
Radiology, Nuclear Medicine and imaging
Neurology (clinical)
Artificial intelligence
business
- Language
- ISSN
- 1120-9976
In this paper an artificial modular system applied to object classification in brain MR images is presented. It consists of two modules based on neural architectures joined in sequence to perform first an image segmentation and then an object classification. For these two steps a Self Organizing Map and a Multilayer Perceptron trained with the Back-Propagation learning rule have been used. The objective of the system is the automatic recognition of the anatomic structures in MR images of the cerebral section passing through the orbits and the visual pathways. To reach this goal we have submitted the two networks to a training phase realized by an unsupervised process for the image segmentation and by a supervised process for regions labelling. This last step has been based on topographic relations supplied by a medical expert. The system has been useful to discriminate 20 different classes of anatomic objects over the considered section. Preliminary results are presented.