Randy Cahya Wihandika, Agus Zainal Arifin, Anny Yuniarti
Mandibular bone is among the bones which mineral density is reduced due to osteoporosis. Hence, dental panoramic gradiographs have become an alternative solution to identify osteoporosis. Previous study has shown that the number of branching differs between normal people and patients with osteoporosis. Another study has proposed an algorithm called COSFIRE which is able to detect branches in retinal vessel images. However, branch structures in trabecular bone differ from that in retinal vessel images. For that reason, the COSFIRE method alone is considered unable to detect such structures. In this study, we propose multiscale mechanism to detect different size of trabecular branches. Morphological structures in the trabecular bone is first enhanced using the line operator method. Then the branches are detected using COSFIRE. Experiment of the branching detection conducted to 20 images yields an accuracy of 95.25% whereas the experiment of the classification step gives the sensitivity, specificity, and accuracy of 0.95122, 0.26315, and 0.55102, respectively. © 2018 Association for Computing Machinery.
Faculty of Computer Science, Brawijaya University, Malang, Indonesia; Faculty of Information and Communication Technology, Sepuluh Nopember Institute of Technology, Surabaya, Indonesia
