分野別セミナー

　Osteoporotic fractures are a health burden worldwide, resulting in reduction of physical activity, increased risk of mortality, and incremental medical cost. Since bone fragility is mainly dependent on skeletal bone mass or bone mineral density (BMD), several types of equipment have been developed to assess skeletal BMD and predict fracture risk from osteoporosis. BMD at the lumbar spine and femoral neck are typically assessed using dual-energy X-ray absorptiometry (DXA). The assessment fee is quite expensive, while the availability of DXA equipment is still too limited to identify a large segment of individuals with undetected osteoporosis. Since the screening system for osteoporosis is insufficient, osteoporosis is still underdiagnosed.
　Several studies on computer-aided system for osteoporosis detection have been conducted at our laboratory. These studies use dental panoramic radiograph because this image is easy to obtain. In general, osteoporosis detection using dental panoramic radiograph is done by measuring the thickness of cortical bone [1-6] or the density of trabecular bone in lower jaw area [7-10]. The dental panoramic radiographs are collected from Indonesian subjects in cooperation with Faculty of Dentistry in Airlangga University and Padjajaran University.
　In general, there are several processes on osteoporosis detection using dental panoramic radiograph, which are image enhancement, segmentation, feature extraction, and classification. Dental panoramic radiograph is low contrast image, so a strategy for image enhancement, especially for contrast adjustment, is needed. We have conducted researches about image enhancement using non-uniform directional filtering [11-13]. This method separates an image into several sub images in the frequency domain based on the orientation of edges in the image. Different enhancement scheme can be applied for different edge orientation.
　There are several parts in jaw area that can be used for osteoporosis detection. Segmentation process is needed to separate those areas (object) from its background. We have conducted several researches about segmentation methods that can be used on low contrast images such as dental panoramic radiographs [14-23]. One of our recent research is about segmentation method on low contrast images using semi-automatic approach [17]. In semi-automatic image segmentation, user marks several regions as the representative sample of object and background cluster. Then regions which have similar properties with the sample will be merged into the cluster of the sample.
　If we use trabecular bone density for osteoporosis detection, another problem arises. Trabecular bone or spongy bone has structure like sponge with many open spaces connected by many small bones. We have developed a method for enhance the structure of trabecular bone and then segment it, called multiscale line operator [9] [10] [24]. This method uses Gaussian Image Pyramid and line operator to detect linear structure of an image with different size. Hence it can detect the structure of trabecular bone which is difficult to observe by human eye.
　The method for feature extraction is different for each feature. Features that can be extracted are including the thickness and density of jaw’s bone. Usually, we calculate the perpendicular line between the upper and lower border of cortical bone for measuring its thickness. For measuring the density of cortical bone, several approaches have been proposed such as using texture features, histogram, or the number of trabecular bone’s pixels.
　The selected features will be used in classifier to determine whether if a patient is at risk of osteoporosis or not. We have conducted several researches for improving the performance of classifier, such as Fuzzy and Naïve Bayes [7]. But usually in our research we use fuzzy neural network for classification process [3-7] or by calculate the correlation to prove the effectiveness of the proposed method [1] [2] [9].
　However, since dental panoramic radiograph is 2-dimensional image taken from front, the thickness or the density of the jaw bone can not be measured as a whole. In addition, the use of X-ray beam on panoramic radiographs can cause noise such as hyoid bone (neck bone) that appears in the jaw area, light projection, and uneven exposure. On the other hand, cone-beam computed tomography (CBCT) is a medical imaging that produces a sequence of 2-dimensional images that can be reconstructed into a 3-dimensional image. Besides, the use of CBCT in osteoporosis detection is still not widely explored.
　In our future work, a computer-aided system to assist osteoporosis assessment will be developed. The system use extracted features from the cone-beam computed tomography (CBCT) images, which taken from maxillofacial (face and jaw) area of the patients. Using the sequence of 2-dimensional images from CBCT, it is expected that the system can measure the volume of the jaw bone.



References



[1]
A. Z. Arifin, A. Asano, A. Taguchi, T. Nakamoto, M. Ohtsuka and K. Tanimoto, "Computer-aided system for measuring the mandibular cortical width on panoramic radiographs in osteoporosis diagnosis," Medical Imaging. International Society for Optics and Photonics, pp. 813-821, 2005.
[2]
A. Z. Arifin, A. Asano, A. Taguchi, T. Nakamoto, M. Ohtsuka, M. Tsuda, Y. Kudo and K. Tanimoto, "Computer-aided system for measuring the mandibular cortical width on dental panoramic radiographs in identifying postmenopausal women with low bone mineral density," Osteoporosis International, vol. 17, no. 5, pp. 753-759, 2006.
[3]
A. Z. Arifin, A. Asano, A. Taguchi, T. Nakamoto, M. Ohtsuka, M. Tsuda, Y. Kudo and K. Tanimoto, "Identification of low bone mineral density based on the mandibular cortex by fuzzy neural network," SCIS & ISIS SCIS & ISIS 2006. Japan Society for Fuzzy Theory and Intelligent Informatics, pp. 1860-1865, 2006.
[4]
A. Z. Arifin, A. Asano, A. Taguchi, T. Nakamoto, M. Ohtsuka, M. Tsuda, Y. Kudo and K. Tanimoto, "A Fuzzy Expert System Design for Diagnosing Osteoporosis Based on Mandibular Cortex Measurement on Dental Panoramic Radiographs," Indonesian Scientific Conference in Japan Proceeding, 2006.
[5]
A. Z. Arifin, A. Asano, A. Taguchi, T. Nakamoto, M. Ohtsuka, M. Tsuda, Y. Kudo and K. Tanimoto, "Developing computer-aided osteoporosis diagnosis system using fuzzy neural network," JACIII, Vols. 1049-1058, p. 11, 2007.
[6]
A. Z. Arifin, A. Asano, A. Taguchi, T. Nakamoto, M. Ohtsuka, M. Tsuda, Y. Kudo and K. Tanimoto, "Use of fuzzy neural network in diagnosing postmenopausal women with osteoporosis based on dental panoramic radiographs," Journal of Advanced Computational Intelligence, vol. 11, no. 8, 2007.
[7]
D. Herumurti, A. Z. Arifin, R. Sulaeman, A. Asano, A. Taguchi, T. Nakamoto and K. Uchimura, "Weighted fuzzy ARTMAP for osteoporosis detection," Proceeding 16th Korea-Japan Joint Workshop on Frontiers of Computer Vision, Seoul, 2007.
[8]
Z. Abidin and A. Z. Arifin, "GENERATION GRAPH WITH RANDOM GRAPH ERDOS ROYI METHOD BY MEDICAL IMAGE TO HELP DIAGNOSES OSTEOPOROSIS," International Conference Bio Medical Engenering, 2008.
[9]
A. Z. Arifin, A. Asano, A. Taguchi, T. Nakamoto, A. Yuniarti, L. R. Dewi and H. Studiawan, "Line strength measurement for trabecular bone analysis of mandible on dental panoramic radiographs," Proceeding of The International Workshop on Advanced Image Technology (IWAIT), 2010.
[10]
A. Z. Arifin, A. Yuniarti, L. R. Dewi, A. Asano, A. Taguchi, T. Nakamoto, A. Razak and H. Studiawan, "Computer-aided diagnosis for osteoporosis based on trabecular bone analysis using panoramic radiographs," Dental Journal (Majalah Kedokteran Gigi), vol. 43, no. 3, pp. 107-112, 2010.
[11]
N. Jawas, A. Z. Arifin, A. Y. Wijaya, A. Yuniarti and W. N. Khotimah, "Non-uniform decimation-free directional filter bank using histogram analysis for image enhancement," 2014 International Conference on Information, Communication Technology and System (ICTS). IEEE, pp. 147-152, 2014.
[12]
A. Z. Arifin, A. Yuniarti, W. N. Khotimah, A. Y. Wijaya, A. Hadi, N. Jawas and E. R. Astuti, "Decimation-Free Directional Filter Banks for Classification and Numbering on Posterior Dental Radiography Using Mesiodistal Neck Detection.," JACIII, vol. 18, no. 4, pp. 649-657, 2014.
[13]
R. Indraswari, A. Z. Arifin, D. A. Navastara and N. Jawas, "Teeth segmentation on dental panoramic radiographs using decimation-free directional filter bank thresholding and multistage adaptive thresholding," 2015 International Conference on Information & Communication Technology and Systems (ICTS). IEEE, pp. 49-54, 2015.
[14]
D. A. Navastara and A. Z. Arifin, "Image Thresholding on Segmentation of Teeth in Dental Panoramic Radiographs," Proceeding of the International Conference on Advanced Computer Science and Information System, 2009.
[15]
N. Dian Tunjung, A. Z. Arifin and R. Soelaiman, "Medical Image Segmentation Using Generalized Gradient Vector Flow and Clifford Geometric Algebra". International Conference on Biomedical Engineering, Surabaya, Indonesia, November 11, 2008.
[16]
K. Umam, A. Z. Arifin, D. A. Navastara, A. Yuniarti, W. N. Khotimah and A. Y. Wijaya, "A Novel Strategy of Differential Evolution Algorithm’s Crossover Operator Based on Graylevel Clusters Similarity for Automatic Multilevel Image Thresholding", International Journal of Control Theory and Applications (IJCTA), No. 2, 2015.
[17]
A. S. Sankoh, A. Z. Arifin and A. Y. Wijaya, "Extracted Pixels Similarity Features (EPSF) using Interactive Image Segmentation Techniques," International Journal of Computer Applications, vol. 132, no. 2, 2016.