Feature Fusion for FDG-PET and MRI for Automated Extra Skeletal Bone Sarcoma Classification

• Main Authors: Baskaran, K., Malathi, R., Thirusakthimurugan, P.
• Format: Conference Proceeding
• Language: English
• Subjects: Bone Sarcoma; Classifiers; Correlation based Feature Selection (CFS); FluorDeoxyGlucose; Grey Level Co-occurrence Matrix (GLCM); Magnetic Resonance Imaging (MRI); Particle Swarm Optimization (PSO); Positron Emission Tomography (FDG-PET); Wavelet
• ISSN: 2214-7853; 2214-7853
• DOI: 10.1016/j.matpr.2017.11.289

The aim of the current work is the evaluation of Positron Emission Tomography (PET) utilizing 18F-Fluoro-2-Deoxy-Dglucose (FDG) when compared to volumetric as well as standard Magnetic Resonance Imaging (MRI) variables for assessing histological responses in bone sarcoma afflicted individuals. The generation of novel composite texture from combining FDG-PET as well as MRI data is examined to see if aggressive tumors could be better identified. For this particular objective, retrospective evaluation was performed on a group of 51 individuals in the same age group. All the patients had pre-treatment FDG-PET as well as MRIs consisting of T1-weighted as well asT2-weighted Fat-Suppression Sequences (T2FS). 9 non-textural features(SUV measures as well as shape attributes) as well as 41 textural features were extracted from the tumor area of distinct as well as fused scans. Extracting features was carried out by Grey Level Co-occurrence Matrix (GLCM), wavelet features. Selecting features was carried out by Correlation based Feature Selection (CFS) as well as Particle Swarm Optimization (PSO). Classification was carried out by K-Nearest Neighbor (KNN), J48 as well as Neural Network (NN).The suggested method attained best classification accuracy.