search
Contact Us
HOME

  • Crossref logo
  • Crossref Similarity Check logo
Journal Search Engine

to

Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer
ISSN : 1225-1577(Print)
ISSN : 2384-0900(Online)
The Korean Journal of Oral and Maxillofacial Pathology Vol.47 No.6 pp.123-131
DOI : https://doi.org/10.17779/KAOMP.2023.47.6.002

Effect of Masking Dental Region on Determining Osteoporosis of Artificial Intelligence on Panoramic Radiographs

Sejin Ahn1), In-Ja Song2), Jae-Seo Lee3), Kyungmin Clara Lee4), Suk-Ja Yoon3)*, Ho-Jun Song5)
1)School of Dentistry, Chonnam National University
2)Department of Nursing, Kwangju Women's University
3)Department of Oral and Maxillofacial Radiology, School of Dentistry, Dental Science Research Institute, Chonnam National University
4)Department of Orthodontics, School of Dentistry, Dental Science Research Institute, Chonnam National University
5)Department of Dental Biomaterials, School of Dentistry, Dental Science Research Institute, Chonnam National University
* Correspondence: Prof. Suk-Ja Yoon, Department of Oral and Maxillofacial Radiology, School of Dentistry, Chonnam National University 33 Yongbongro Bukgu Gwangju, 61186 South Korea Tel: +82-62-530-5680 Email: yoonfr@chonnam.ac.kr
October 19, 2023 October 26, 2023 December 15, 2023

Abstract


This study aimed to investigate which areas AI is sensitive when inputting panoramic radiographs with dental area masked and when inputting unmasked ones. Therefore, the null hypothesis of this study was that masking dental area would not make a difference in the sensitive areas of osteoporosis determination of AI. For this study 1165 female(average age 48.4 ± 23.9 years) from whom panoramic radiographs were taken were selected. Either osteoporosis or normal should be clearly defined by oral and maxillofacial radiologists. The panoramic radiographs from the female were classified as either osteoporosis or normal according to the mandibular inferior cortex shape. VGG-16 model was used to get training, validating, and testing to determine between osteoporosis or normal. Two experiments were performed; one using unmasked images of panoramic radiographs, and the other using panoramic radiographs with dental region masked. In two experiments, accuracy of VGG-16 was 97.9% with unmasked images and 98.6% with dental-region-masked images. In the osteoporosis group, the sensitive areas identified with unmasked images included cervical vertebrae, maxillary and mandibular cancellous bone, dental area, zygomatic bone, mandibular inferior cortex, and cranial base. The osteoporosis group shows sensitivity on mandibular cancellous bone, cervical vertebrae, and mandibular inferior cortex with masked images. In the normal group, when unmasked images were input, only dental region was sensitive, while with masked images, only mandibular cancellous bone was sensitive. It is suggestive that when dental influence of panoramic radiographs was excluded, AI determined osteoporosis on the mandibular cancellous bone more sensitively.


Osteoporosis , Mandible , Panoramic Radiography , Artificial Intelligence

인공지능의 파노라마방사선영상에서 골다공증 판정에 미치는 치아영역 마스킹의 효과

안세진1), 송인자2), 이재서3), 이경민4), 윤숙자3)*, 송호준5)
1)전남대학교 치의학전문대학원
2)광주여자대학교 간호학과
3)전남대학교 치의학전문대학원 구강악안면방사선학교실, 치의학연구소
4)전남대학교 치의학전문대학원 교정학교실, 치의학연구소
5)전남대학교 치의학전문대학원 치과재료학교실, 치의학연구소

초록

    Ⅰ. INTRODUCTION

    Osteoporosis is a disease characterized by decreased bone density, destruction of bone tissue, abnormalities in bone microstructure, decreased bone strength, and fractures 1). Dual-energy X-ray absorptiometry(DXA) is a proven technology and is generally applied to the hip, vertebrae, and forearm. The standards of the world health organization( WHO) can be applied to DXA, but the disadvantages include high cost and high dose of radiation exposure. Among osteoporosis patients with fractures, there are no symptoms in the early stages, so the number of osteoporosis patients diagnosed with DXA is very small1,2). Meanwhile, studies have reported that reading mandible inferior cortex shape on dental panoramic radiography shows similar sensitivity and specificity as DXA in diagnosing osteoporosis3,4-15). Panoramic radiography is a conventional radiography for routine checks in dental clinics, and has the advantage of being more advantageous than DXA in terms of cost and exposure dose16).

    Recently, studies on the determination of osteoporosis on panoramic radiography using artificial intelligence(AI) have been reported17-19). Convolutional neural network(CNN) was first applied to panoramic radiographs in determining osteoporosis 17).

    The ability of AI has been reported to be very excellent in determining osteoporosis on panoramic radiographs. Depending on the research conditions such as study subjects, age groups, limiting region of interest(ROI), etc., sensitivity was reported as 91.5-100%, specificity 52.9-99.0%, and accuracy 84.84-99.25%17-24). Oral and Maxillofacial Radiologists made diagnosis by reading the mandible inferior cortex shape, but AI was sensitive elsewhere including maxillary and mandibular cancellous bone18,22-24). Recently, a study examined AI's ability to determine osteoporosis by inputting the entire panoramic radiograph without limiting ROI. In this study, the cervical vertebrae were included in the sensitive areas of AI in the osteoporosis group. On the other hand, the dental area was included in the sensitive areas in the normal group24). Vertebrae showed osteoporosis in the osteoporosis group, and this finding might reveal that AI has the ability to determine osteoporosis. However, it was an unexpected result that dental area affected the determination of osteoporosis.

    Therefore, this study aimed to determine how dental area affects the determination of osteoporosis. We wanted to study which areas AI is sensitive when inputting panoramic radiographs with dental area masked and when inputting unmasked ones. Therefore, the null hypothesis of this study was that masking dental area would not make a difference in the sensitive areas of osteoporosis determination of AI.

    Ⅱ. MATERIALS and METHODS

    1. Panoramic radiographs

    This study was implemented by the approval of the Chonnam National University institutional review board (CNUDH-2017-014)17). This study was one of a serial study that has investigated the effectiveness of osteoporosis diagnosis on panoramic radiographs using CNN17-19,22-24). For this study 1165 panoramic radiographs were used which were taken from female(mean age 48.4 ± 23.9 yrs) who visited Chonnam National University Dental Hospital between 2008 and 2017. Panoramic radiographs were obtained with Kodak 8000C(Carestream Health Inc,. Rochester, NY). Panoramic radiographs should be clearly diagnosed as osteoporotic or normal. All panoramic radiographs were classified into osteoporosis and normal by two Oral and Maxillofacial Radiologists with more than 10 years of experience.

    Osteoporosis was determined according to mandibular inferior cortex shape using the method of Klemetti et al25). The mandible inferior cortex was classified into three, C1, C2, and C3. The cortex with even radiopacity and sharp border was C1. The cortex with mild to moderate erosion or semilunar defects was C2. The cortex with clear porosity was C3. C1 was categorized as normal, while C2 and C3 as osteoporosis. Even if it was C1, it was diagnosed to be osteoporotic when the cortex was very thin26). The final diagnosis was made by the agreement of the two Oral and Maxillofacial Radiologists, excluding the patients from the study subjects in whom the diagnosis was not clear or the two Oral and Maxillofacial Radiologists disagreed17).

    2. Deep learning model

    VGG-16, a type of deep learning CNN, was used. VGG-16 consists of 16 layers and has an architecture in which a convolution layer and a max pooling layer are stacked. The output is derived through a fully connected layer after passing through layered structures. This study used a global average pooling layer to apply gradientweighted class activation mapping(Grad-CAM). Stochastic gradient descent was used as an optimizer for training, and the settings were set to learning rate 0.005 and momentum 0.9. In the case of loss-entropy, categorical cross entropy was used, and 100 epochs were performed27).

    Among a total of 1,165 female, 547(73.9 ± 7.75 yrs) had osteoporosis and 619(27.7 ± 5.37 yrs) had normal disease, which was diagnosed on panoramic radiography. The sets were classified for AI training, validation, and testing. As training sets, 383 female(77.8 ± 4.7 yrs) with osteoporosis and 433(29.8 ± 4.7 yrs) normal were used. As validation sets, 82(67.9 ± 1.60 yrs) with osteoporosis and 92(21.9 ± 2.96 yrs) normal were used, and as testing sets, 70(59.8 ± 3.32 yrs) with osteoporosis and 74(22.5 ± 0.50 yrs) normal were used. The panoramic radiographs, which are input to AI, were used for two experiments with unmasked images and masked images in the dental area. The panoramic radiographs on which dental area was masked with a black rectangle were defined as masked images, and those without masking as unmasked images. Experiment 1 and 2 were performed with the same training, validating, and testing sets. Experiment 1 used unmasked images to determine osteoporosis or normal, and Experiment 2 used masked images( Figure 1, Table 1).

    3. Evaluation of deep learning prediction

    Accuracy was obtained to evaluate the results obtained after learning based on the above conditions. In addition, the decision-making process by a deep learning model was analyzed by forming a heatmap image using Grad-CAM to display the sensitive areas that had a major influence on each panoramic radiolograph. The heatmap was displayed in a color gradiation from blue(low) to red(high) depending on the degree of influence in determining osteoporosis and normal using VGG-1627).

    Ⅲ. RESULTS

    The training sets of VGG-16 consisted of panoramic radiographs of 383 female (77.8 ± 4.7 yrs) with osteoporosis and 433 normal female (29.8 ± 4.7 yrs). The testing sets consisted of 70(59.8 ± 3.32 yrs) with osteoporosis and 74(22.5 ± 0.50 yrs) normal. In the experiment 1 with the unmasked images, VGG-16 determined all 74 out of 74 normal female to be normal, while 67 out of 70 osteoporotic female were determined as osteoporosis. The accuracy was 97.9%. In the experiment 2 with the masked images, 73 out of 74 normal female were determined to be normal, and 69 out of 70 osteoporosis female were determined as osteoporosis. Accuracy was 98.6%. The sensitive areas were checked on panoramic radiographs on which VGG-16 accurately determined osteoporosis and normal. Heatmap images using Grad-CAM were displayed in a gradiation from blue to red, and the area that has the great influence on determining the osteoporosis in panoramic radiographs is shown in red. On unmasked images of the osteoporosis group, there were a total of 154 sensitive areas marked in red. Among 154, cervical vertebrae was 39(25.3%), maxillary cancellous bone 31(20.1%), mandibular cancellous bone 28(18.2%), dental area 24(15.6%), zygomatic bone 18(11.7%), mandibular inferior cortex 9(5.84%), and cranial base 5(3.25%). The normal group was all(100%) sensitive in the dental area. On the masked images of the osteoporosis group, there were a total of 86 sensitive areas marked in red. Among these, mandibular cancellous bone was 68(79%), cervical vertebrae 12(14%), and mandibular inferior cortex 6(7%). The normal group was all(100%) sensitive in the mandibular cancellous bone(Figure 2-3, Table 2-4).

    Ⅳ. DISCUSSION

    This study aimed to determine how dental area affects the determination of osteoporosis. We wanted to study which areas AI is sensitive when inputting panoramic radiographs with the dental area masked and when inputting unmasked ones. Therefore, the null hypothesis of this study was that masking the dental area would not make a difference in the sensitive areas of osteoporosis determination of AI.

    The most common sequela arising from osteoporosis is ‘fracture’. The most common fractures include the vertebrae, femur, and wrist. Fractures occur unnoticed at first and are not diagnosed because there are few symptoms, and in many cases, the first sign is a low-energy fracture of the vertebrae. Fractures that occur in the body significantly reduce the patient's quality of life. In general, the mortality rate within 1 year after femoral fracture is 15-20%28-32). Early diagnosis of osteoporosis is very important for the healthy life of the elders.

    Panoramic radiography is a simple imaging method that can be helpful in early diagnosis of osteoporosis. Klemetti et al.25) first attempted to study osteoporosis by classifying the shape of the mandible inferior cortex from a panoramic radiograph. Since then, various researchers have conducted studies comparing mandible cortex shape with DXA to determine whether it is useful in diagnosing osteoporosis, and it has been confirmed that it shows up to 93% sensitivity and 89% specificity4-15). When diagnosing osteoporosis, using the shape of the mandible inferior cortex, attention should be paid to whether the trabecular bone tail is connected to the mandibular inferior cortex. Also, in case of C1, if it is very thin, it should be classified as osteoporosis 26). In this study osteoporosis was diagnosed by taking these points into consideration.

    CNN, a field of deep learning included in the AI category, is commonly used in the medical field to classify and detect diseases to obtain relevant images. Research into the application of computer-based diagnosis using this is rapidly expanding, and it is expected that more efficient and faster diagnosis will be possible33).

    This study is one of a serial study that has investigated the effectiveness of osteoporosis diagnosis on panoramic radiographs using CNN17-19,22-24). Rather than confirming the accuracy of osteoporosis judgment, it checked whether significant results were obtained according to masking dental area. The shape of the mandible inferior cortex was read in diagnosing osteoporosis by Oral and Maxillofacial Radiologists. However, the sensitive areas of CNN in determining osteoporosis included other areas including cancellous bone18,22-24). Recently, An et al.24) used entire panoramic radiographs for testing CNN for osteoporosis, for osteoporosis is a disease of whole body skeleton. In osteoporosis group, 70.8% were sensitive in the cervical vertebrae, 72.2% were sensitive in the mandible inferior cortex, 33.3% were sensitive in the mandible cancellous bone, and 20.8% were sensitive in the maxillary cancellous bone. The cranial base, zygomatic bone, and dental area were also sensitive. In normal group, 89.9% were sensitive in the dental area, and in addition, the mandible inferior cortex, mandibular and maxillary cancellous bone were sensitive. Vertebrae is a bone where osteoporosis frequently occurs, and CNN showed that it accurately diagnosed osteoporosis on the entire panoramic radiograph. On the other hand, it was an unconvincing finding that dental areas unrelated to bone were read sensitively. Therefore, there was a need to study the effect of dental area on osteoporosis diagnosis24).

    In this study, two experiments were performed by inputting masked and unmasked images of the dental area into CNN, respectively. Using unmasked images, accuracy was 97.9%, while using masked images, accuracy was 98.6%. There were differences in the sensitive areas marked in red between masked and unmasked images. Sensitive areas in the osteoporosis group on unmasked images included cervical vertebrae (25.3%), maxillary cancellous bone (20.1%), and mandible inferior cortex (5.84%), mandibular cancellous bone (18.2%), zygomatic bone (11.7%), cranial base (3.25%), dental area (15.6%), etc. Cervical vertebrae were sensitive at high frequency, which was similar to the study by An et al17). On masked images, the mandibular cancellous bone was sensitive with 79%, which was higher than in other areas in the osteoporosis group. The normal group was all(100%) sensitive in the mandibular cancellous bone. These results suggest that masking dental area affect the sensitive areas of osteoporosis readings. As a result, the null hypothesis of this study was rejected that masking the dental area in this study will not make any difference in the sensitive areas of osteoporosis determination(Figure 2-3, Table 2-4).

    Among the factors contributing to bone loss, the most important is cancellous bone, followed by cortex34). Cancellous bone is less dense and has more flexible properties than cortex. It can be seen that cancellous bone has a faster loss rate and is more sensitive than cortex35). Using this characteristic, a method of diagnosing osteoporosis using the density of lumbar vertebrae, which are rich in cancellous bone, is widely used36). The mandibular cancellous bone responded more sensitively when dental area was masked in CNN compared to unmasked images(Table 4).

    Osteoporotic change of cancellous bone was more sensitive by CNN because CNN is able to recognize the subtle bone change more accurately than human eyes. When Oral and Maxillofacial Radiologists diagnose osteoporosis, they focus on cortex change because interpreting cortex change is easier than cancellous bone with naked eyes. Human eyes can distinguish only about 30 gray levels or less, while CNN can distinguish a wide range of gray levels. The contrast resolution of computer is described as dynamic range. The dynamic range of computer system is identified by the bit capacity of each pixel. Recently computer system provides more than 14-bit dynamic range, which is 16,384 shades of gray, which explains the difference of reading ability between Oral and Maxillofacial Radiologists and CNN37,38).

    There were limitations of this study. The study sample was not large and there was a large age difference between osteoporosis and normal groups. Only the dental area was not masked in detail during the masking process of the panoramic radiolographs. The determination of osteoporosis using only panoramic radiographs without DXA was also a limitation.

    In conclusion, in this study, after masking the dental area, the sensitive area was marked with 79.8% in the mandibular cancellous bone in the osteoporosis group, and in the normal group, only the mandibular cancellous bone was sensitive with 100%. It is suggestive that when dental influence of panoramic radiographs was excluded, AI determined osteoporosis on the mandibular cancellous bone more sensitively.

    Figure

    KAOMP-47-6-123_F1.gif

    VGG-16 architecture for unmasked and masked images in determining osteoporosis or normal

    KAOMP-47-6-123_F2.gif

    Unmasked(1) and masked(2) Images with sensitive area in red on 3 female A, B, and C with osteoporosis using Grad-Cam Image.

    KAOMP-47-6-123_F3.gif

    Unmasked(1) and masked(2) Images with sensitive area in red on 2 normal female A and B using Grad-Cam Image.

    Table

    Study subjects for this study (Total 1165 female)

    The results of experiment 1 in determining osteoporosis using unmasked images.

    CNN: convolutional neural network; OMFR: Oral and Maxillofacial Radiologists

    The results of experiment 2 in determining osteoporosis using masked images.

    CNN: convolutional neural network; OMFR: Oral and Maxillofacial Radiologists

    Sensitive areas on the images on which osteoporosis or normal was correctly determined by VGG-16 in experiment 1 and 2.

    N : the number of panoramic radiographs which were correctly determined either osteoporosis or normal

    Reference

    1. World Health Organization: WHO scientific group on the assessment of osteoporosis at primary health care level. Summary Meeting Report, Brussels, Belgium. World Health Organization 2007:1-17.
    2. Park YS: Diagnosis and treatment of osteoporosis. J Korean Med Assoc 2012;55:1083-1094.
    3. Taguchi A: Triage screening for osteoporosis in dental clinics using panoramic radiographs. Oral Dis. 2010;16:316-327.
    4. Drozdzowska B, Pluskiewicz M, Tarnawska B: Panoramicbased mandibular indices in relation to mandibular bone mineral density and skeletal status assessed by dual energy X-ray absorptiometry and quantitative ultrasound. Dentomaxillofac Radiol 2002;31: 361-367.
    5. Yaşar F, Akgünlü F: The differences in panoramic mandibular indices and fractal dimension between patients with and without spinal osteoporosis. Dentomaxillofac Radiol 2006;35:1-9.
    6. Vlasiadis K. Z, Skouteris C. A, Velegrakis G. A, Fragouli I, Neratzoulakis J. M, Damilakis J, Koumantakis E. E: Mandibular radiomorphometric measurements as indicators of possible osteoporosis in postmenopausal women. Maturitas 2007;58:226-235.
    7. Taguchi A, Suei Y, Sanada M, Ohtsuka M, Nakamoto T, Sumida H, Ohama K, Tanimoto K: Validation of dental panoramic radiography measures for identifying postmenopausal women with spinal osteoporosis. Am J Roentgenol 2004;183:1755-1760.
    8. Taguchi A, Tsuda M, Ohtsuka M, Kodama I, Sanada M, Nakamoto T, Inagaki K, Noguchi T, Kudo Y, Suei Y, Tanimoto K, Bollen A. M: Use of dental panoramic radiographs in identifying younger postmenopausal women with osteoporosis. Osteoporos Int. 2006;17: 387-394.
    9. Taguchi A, Ohtsuka M, Tsuda M, Nakamoto T, Kodama I, Inagaki K, Noguchi T, Kudo Y, Suei Y, Tanimoto K: Risk of vertebral osteoporosis in post-menopausal women with alterations of the mandible. Dentomaxillofac Radiol 2007;36:143-148.
    10. Horner K, Karayianni K, Mitsea K, Berkas L, Mastoris M, Jacobs R, Lindh C, Stelt P, Marjanovic E, Adams J, Pavitt S, Devlin H: The mandibular cortex on radiographs as a tool for osteoporosis risk assessment: the OSTEODENT Project 2007;10:138-146.
    11. Taguchi A, Asano A, Ohtsuka M, Nakamoto T, Suei Y, Tsuda M, Kudo Y, Inagaki K, Noguchi T, Tanimoto K, Jacobs R, Klemetti E, White S. C, Horner K: Observer performance in diagnosing osteoporosis by dental panoramic radiographs: results from the osteoporosis screening project in dentistry (OSPD). Bone 2008;43:209-2013.
    12. Nakamoto T, Taguchi A, Ohtsuka M, Suei Y, Fujita M, Tanimoto K, Tsuda M, Sanada M, Ohama K, Takahashi J, Rohlin M: Dental panoramic radiograph as a tool to detect postmenopausal women with low bone mineral density: untrained general dental practitioners’ diagnostic performance. Osteoporos Int 2003;14:659-664.
    13. Halling A, Persson G. R, Berglund J, Johansson O, Renvert S: Comparison between the Klemetti index and heel DXA BMD measurements in the diagnosis of reduced skeletal bone mineral density in the elderly. Osteoporos Int 2005;16:999-1003.
    14. White S. C, Taguchi A, Kao D, Wu S, Service S. K, Yoon D, Suei Y, Nakamoto T, Tanimoto K: Clinical and panoramic predictors of femur bone mineral density. Osteoporos Int 2005;16:339-346.
    15. Sutthiprapaporn P, Taguchi A, Nakamoto T, Ohtsuka M, Mallick P.C, Tsuda M, Kodama I, Kudo Y, Suei Y, Tanimoto K: Diagnostic performance of general dental practitioners after lecture in identifying post-menopausal women with low bone mineral density by panoramic radiographs. Dentomaxillofac Radiol 2006;35:249-252.
    16. GauR B, Chaudhary A, Wanjari PV, Sunil MK, Basavaraj P: Evaluation of panoramic Radiographs as a Screening Tool of Osteoporosis in Post Menopausal Women. A Cross Sectional Study. J Clin Diagn Res 2013;7:2051-2055.
    17. Lee JS, Adhikari S, Liu L, Jeong HG, Kim H, Yoon SJ: Osteoporosis detection in panoramic radiographs using a deep convolutional neural network-based computer-assisted diagnosis system: a preliminary study. Dentomaxillofac Radiol 2019;48:20170344.
    18. Bae SY, Song IJ, Kim H, Adhikari S, Lee JS, Yoon SJ: AlexNet in Determining Osteoporosis on Dental Panoramic Radiograph. Korean J Oral Maxillofac Pathol. 2021;45:189-196.
    19. Kim JY, Lee JS, Kang BC, Kim HS, Adhikari S, Liu L, Yoon SJ: Effect of Training and Testing Condition of Convolutional Neural Network on evaluating Osteoporosis. Korean J Oral Maxillofac Pathol 2019;43:73-80.
    20. Tassoker M, Öziç MÜ, Yuce F: Comparison of five convolutional neural networks for predicting osteoporosis based on mandibular cortical index on panoramic radiographs. Dentomaxillofac Radiol 2022;51:20220108.
    21. Lim YG, Lee EJ, Lee JS, Kang BC, Adhikari S, Kim H: Differences in Osteoporosis Readings on Dental Panoramic Radiographs according to Convolutional Neural Network Test Data. Korean J Oral Maxillofac Pathol 2019;43:103-109.
    22. Song J, Song IJ, Kim H, Adhikari S, Lee JS, Yoon SJ: Mandibular Cortical Thinning Detection of Deep Convolutional Neural Network on Panoramic Radiographs. Korean J Oral Maxillofac Pathol 2021;45:157-164.
    23. Lee SA, Park BJ, Kim H, Adhikari S, Lee EJ, Lee JS: Sensitive Area of Artificial Intelligence in Interpreting Osteoporosis on Panoramic Radiograph. Korean J Oral Maxillofac Pathol 2021;45:59-65.
    24. An JI, Song IJ, Song HJ, Park BJ, Lee JS ,Yoon SJ: Ability to Determine Osteoporosis of the ConvolutionalNeural Network for the Entire Area of Panoramic
    25. Klemetti E, Kolmakov S, Kröger H: Pantomography in assessment of the osteoporosis risk group. European J Oral Sciences 1994;102:68–72.
    26. Taguchi A: Panoramic radiographs for identifying individuals with undetected osteoporosis. Japanese Dental Science Review 2009;45:109-120.
    27. Ferguson M, Ak R, Lee Y. T, Law K H: Automatic localization of casting defects with convolutional neural networks. 2017 IEEE International Conference on Big Data. IEEE. 2017;1726-735.
    28. Mackey D, C, Lui L. Y, Cawthon P. M, Bauer D. C, Nevitt M. C, Cauley J. A, Hillier T. A, Lewis C. E, Barrett-Connor E, Cummings S. R: Study of Osteoporotic Fractures (SOF) and Osteoporotic Fractures in Men Study (MrOS) Research Groups. High-trauma fractures and low bone mineral density in older women and men. JAMA. 2007;298: 2381-2388.
    29. Sukegawa S, Fujimura A, Taguchi A, Yamamoto N, Kitamura A, Goto R, Nakano K, Takabatake K, Kawai H, Nagatsuka H, Furuki Y: Identification of osteoporosis using ensemble deep learning model with panoramic radiographs and clinical covariates. Sci Rep 2022;12:6088.
    30. LeBoff MS, Greenspan SL, Insogna KL, Lewiecki EM, Saag KG, Singer AJ, Siris ES: The clinician’s guide to prevention and treatment of osteoporosis. Osteoporo Int 2022;33:2049-2102.
    31. Hearthcare Bigdata hub, Health care, Statistical information :http://opendata.hira.or.kr/op/opc/olapMfrnIntrsIlnsInfo.do
    32. Marcucci G, Brandi ML: Rare causes of osteoporosis. Clin Cases Miner Bone Metab 2015;12:151-156.
    33. Kwon O, Yong TH, Kang SR, Kim JE, Huh KH, Heo MS, Lee SS, Choi SC, Yi WJ: Automatic diagnosis for cysts and tumors of both jaws on panoramic radiographs using a deep convolution neural network. Dentomaxillofac Radiol 2020;49:20200185.
    34. Chen H, Zhou X, Fujita H, Onozuka M, Kubo KY: Age-related changes in trabecular and cortical bone microstructure. Int J Endocrinol 2013;2013:213234.
    35. Han ZH, Palnitkar S, Rao DS, Nelson D, Parfitt AM: Effect of ethnicity and age or menopause on the structure and geometry of iliac bone. J Bone Miner Res 1996;11:1967-1975.
    36. Hong SB: Treatment and Diagnosis of Osteoporosis. J Korean Neurol Assoc. 2017;35:20-24.
    37. Bushong SC: Radiologic science for technologists. Mosby. Elsevier. 2008;9:427-435.
    38. Kimpe T, Tuytschaever T: Increasing the number of gray shades in medical display systems-how much is enough?. J Digit Imaging 2007;20:422-432.
    Copyright © Korean Association of Oral and Maxillofacial Pathology. All rights reserved.
    오늘하루 팝업창 안보기 닫기