MRI Texture-Based Classification of Dystrophic Muscles. A Search for the Most Discriminative Tissue Descriptors
The study assesses the usefulness of various texture-based tissue descriptors in the classification of canine hindlimb muscles. Experiments are performed on T2-weighted Magnetic Resonance Images (MRI) acquired from healthy and Golden Retriever Muscular Dy
- PDF / 337,624 Bytes
- 13 Pages / 439.37 x 666.142 pts Page_size
- 96 Downloads / 138 Views
2
Faculty of Computer Science, Bialystok University of Technology, Wiejska 45a, 15-351 Bialystok, Poland {d.duda,m.kretowski}@pb.edu.pl Institute of Myology, Nuclear Magnetic Resonance Laboratory, Paris, France 3 CEA, I2BM, MIRCen, NMR Laboratory, Paris, France
Abstract. The study assesses the usefulness of various texture-based tissue descriptors in the classification of canine hindlimb muscles. Experiments are performed on T2-weighted Magnetic Resonance Images (MRI) acquired from healthy and Golden Retriever Muscular Dystrophy (GRMD) dogs over a period of 14 months. Three phases of canine growth and/or dystrophy progression are considered. In total, 39 features provided by 8 texture analysis methods are tested. Features are ranked according to their frequency of selection in a modified Monte Carlo procedure. The top-ranked features are used in differentiation (i) between GRMD and healthy dogs at each phase of canine growth, and (ii) between three phases of dystrophy progression in GRMD dogs. Three classifiers are applied: Adaptive Boosting, Neural Networks, and Support Vector Machines. Small sets of selected features (up to 10) are found to ensure highly satisfactory classification accuracies. Keywords: Golden Retriever Muscular Dystrophy (GRMD) · Duchenne Muscular Dystrophy (DMD) · Texture analysis · Feature selection · Classification · MRI T2
1
Introduction
Duchenne Muscular Dystrophy (DMD) is a genetic disorder affecting approximately 1 in 3,600 boys worldwide [1]. It is caused by the absence of dystrophin, a protein that plays an essential role in supporting fiber strength, mainly in the skeletal and cardiac muscles. In affected individuals dystrophin is not synthesized normally, which results in progressive muscle degeneration. This leads to permanent progressive disability (decreased mobility, deformities, cardiomyopathy, and respiratory failure) and premature death [2]. No treatment can reverse the fatal muscle destruction and there is still no effective cure for DMD. c IFIP International Federation for Information Processing 2016 Published by Springer International Publishing Switzerland 2016. All Rights Reserved K. Saeed and W. Homenda (Eds.): CISIM 2016, LNCS 9842, pp. 116–128, 2016. DOI: 10.1007/978-3-319-45378-1 11
MRI Texture-Based Classification of Dystrophic Muscles
117
The Golden Retriever Muscular Dystrophy (GRMD) canine model is the most widely used in research on potential treatment of DMD in humans. It mimics the human DMD model in many aspects [3]. Important information about the progression of the disease and/or its response to therapy can be obtained, in an atraumatic manner, using Magnetic Resonance Imaging (MRI). However, the interpretation of image content is not a trivial task. Great hope is placed in computer-aided image recognition methods, especially those based on texture analysis (TA) [4,5]. The use of appropriately selected textural features in the tissue differentiation process could reduce the need for invasive diagnostic methods, such as those involving needle biopsies, which ca
Data Loading...
