Citra Radiografi Panoramik pada Tulang Mandibula untuk Deteksi Dini Osteoporosis dengan Metode Gray Level Cooccurence Matrix (GLCM)

Abstract

Osteoporosis is one of the degenerative diseases associated with aging, which is apparent from changes in trabecular structure and decreased bone mineral density (BMD) The aim of this study was to obtain a panoramic image quantification method on a region of interest (ROI) to determine the BMD. This study used an ROI (80x80 pixels) of the mandibular condyle for image quantification. The study was performed at the Department of Radiology, Faculty of Dentistry, Padjadjaran University during the period of October to December 2013. A texture analysis approach was applied using the principles of gray level co-occurence matrix (GLCM). The design of image quantification consisted of training and testing stages. The training stage was performed through 9 training data on the subjects of post-menopausal women between 52–73 years old . Data from the lumbar vertebrae BMD DEXA was used as a reference in the classification stage using a support vector machine (SVM) with kernel function multilayer perceptron. The testing used 14 test data from subjects which were not used for training data. The results showed that for the normal and osteoporotic class classification using SVM the accuracy was 85.71%, sensitivity (true positive rate) was 90.91%, and specificity (true negative rate) was 66.67%. The best feature recognition was obtained using a combination of feature contrast, correlation, energy, and homogeneity as inputs for SVM classification. In conclusion, analysis of the trabecular texture using dental panoramic image produced by gray level co-occurance matrix (GLCM) method can be useful for early detection of osteoporosis.

Keywords

• Grey level co-occorance matrix (GLCM)
• panoramic
• osteoporosis