利用全口牙根尖影像特徵值對牙周病之定量分析

牙周病是常見的牙科疾病，由於多數人未有定期做牙齒檢查的習慣，加上牙周病不一定會引起牙齒疼痛，易使人疏忽，往往是因為非牙周病原因就診才發現罹患牙周病。本研究目的是比較牙周病患全口根尖影像，藉由影像處理技術擷取牙齒影像特徵值，利用相關分析、邏輯斯特迴歸(Logistic Regression, LR)和ROC(Receiver Operating Characteristic)曲線分析牙周病臨床症狀、BMI(Body Mass Index)、年紀和影像特徵值間的關係。經由Kappa一致性統計等統計方法進行模型與重要顯著影像特徵分類效能評估，藉此方式可預測病人罹患牙周病之機率。藉由LR模型設定切點為0.5，以及影像平均值切點為75.725，共兩種方法進行影像分類。LR與影像平均值對於影像分類之準確度和一致性係數分別為(99.2%, 0.964)與(86.8%, 0.55)，代表LR具有較佳的檢測能力。與罹患牙周病有關的參數為年紀、BMI和影像熵值、影像平均值、影像標準差，其中以影像平均值為關鍵性特徵參數。LR模型一致性係數高達0.964，代表由LR進行分類之預測結果與臨床診斷結果相符性幾乎一致，可提供醫師診斷牙周病患者一個參考性定量資訊。

關鍵字

牙周病； ROC曲線； Kappa一致性統計量

並列摘要

Periodontal disease is a common dental disease. Most people had been negligent getting a thorough dental check up regularly because periodontal disease can not always cause pain. Often due to other reasons for treatment and found that suffering from periodontal disease. The purpose of this study is used of full-mouth periapical image and captures the imaging features by image processing techniques from periodontal disease patients. The use of Pearson, Logistic Regression (LR) and ROC (Receiver Operating Characteristic) curve analysis to explore the relationship among clinical symptoms of periodontal disease, BMI (Body Mass Index), age and the imaging features. The classification performance with important image features assesses by Kappa consistency of statistical methods, thereby to estimate the probability of a patient suffering from periodontal disease. By LR model set cut-off point for the 0.5, and image average cut-off point for the 0.5,75.725 image classification of the two methods, classification of the LR and image average, accuracy and consistency coefficient was ( 99.2% , 0.964 ) and ( 86.8% , 0.55 ). LR represents a good detection capabilities. Suffering from periodontal disease-related parameters for the age, BMI, image entropy, image average, image standard deviation, which is a key feature of average parameters. LR model consistency factor of up to 0.964, represented by LR to predict the outcome of the classification match is almost consistent with the clinical findings, the result of quantitative information could offer doctors a trustable reference to diagnose periodontal disease.

並列關鍵字

Logistic Regression ； Periodontal Disease ； ROC Curve ； Kappa Statistics

參考文獻

1.Corbet, E. F., & Leung, W. K. (2011). Epidemiology of periodontitis in the Asia and Oceania regions. Periodontology 2000, 56(1), 25-64.

2.Mah, P., McDavid, W. D., & Dove, S. B. (2011). Quality assurance phantom for digital dental imaging. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, and Endodontology, 112(5), 632-639.

4.Ferrús-Torres, E., Gargallo-Albiol, J., Berini-Aytés, L., & Gay-Escoda, C. (2009). Diagnostic predictability of digital versus conventional panoramic radiographs in the presurgical evaluation of impacted mandibular third molars. International Journal of Oral and Maxillofacial Surgery, 38(11), 1184-1187.

5.Koong, B. (2012). The basic principles of radiological interpretation. Australian Dental Journal, 57(s1), 33-39.

6.Hegde, S., & Shetty, S. R. (2012). Radiological features of familial Gorlin-Goltz syndrome. Imaging Science in Dentistry, 42(1), 55-60.

國際替代計量

利用全口牙根尖影像特徵值對牙周病之定量分析

全文下載

主題瀏覽