Department of Electrical and Computer Engineering, São Carlos School of Engineering, University of São Paulo, Av. Trab. São Carlense, 400, São Carlos, São Paulo, 13566-590, Brazil.
Department of Oral Diagnosis, Division of Oral Radiology, Piracicaba Dental School, University of Campinas, Av. Limeira, 901, Piracicaba, São Paulo, 13414-903, Brazil.
Clin Oral Investig. 2021 Aug;25(8):5077-5085. doi: 10.1007/s00784-021-03820-z. Epub 2021 Feb 5.
To present an image processing framework to improve the detection of vertical root fractures (VRFs) in digital periapical radiography.
Thirty endodontically treated human teeth (15 of them fractured with a metal post inserted into them, and 15 for the control) were enclosed in a dry mandible and radiographed individually. The proposed framework was applied to the raw data, as a preprocessing step, and was composed of four stages: geometric adjustment and negative, denoising, adaptive contrast enhancement, and gamma correction. The contrast-to-noise ratio (CNR) and sharpness of the image's VRF region were used for the objective evaluation of the method. In addition, five examiners evaluated the original and enhanced images, using a 5-point scale to assess confidence.
The objective results showed that the proposed framework increased the CNR of the VRF region by 173% compared to the standard preprocessing method provided by the detector's manufacturer. The results found by the human observers indicated that the area under the curve (AUC) and sensitivity of the diagnosis of VRF significantly increased by 4% and 17% (p ≤ 0.05), respectively, when the examiners evaluated the image with the proposed method concomitantly with the image available in the commercial software. However, the specificity was reduced.
The proposed image processing framework can be used as an additional tool to that provided by the manufacturer to increase the sensitivity and AUC of the diagnosis of VRF.
The proposed method can be easily used in clinical practice to aid VRF detection, since it does not incur high computational costs and does not increase the radiation dose applied to the patient.
提出一种图像处理框架，以提高数字根尖射线照相中垂直根折 (VRF) 的检测能力。
将 30 个人工根管治疗后的牙齿（其中 15 颗有金属桩插入）和 15 颗对照牙封闭在干燥的下颌骨中，单独进行放射照相。所提出的框架应用于原始数据，作为预处理步骤，由四个阶段组成：几何调整和负片、去噪、自适应对比度增强和伽马校正。图像 VRF 区域的对比噪声比 (CNR) 和锐度用于方法的客观评估。此外，五位评估员使用 5 分制评估原始和增强图像，以评估信心。
客观结果表明，与探测器制造商提供的标准预处理方法相比，所提出的框架将 VRF 区域的 CNR 提高了 173%。人类观察者的结果表明，当评估员同时使用该方法评估图像和商业软件中提供的图像时，诊断 VRF 的曲线下面积 (AUC) 和灵敏度分别显著提高了 4%和 17%（p ≤ 0.05），但特异性降低。
所提出的图像处理框架可用作制造商提供的附加工具，以提高 VRF 诊断的灵敏度和 AUC。
所提出的方法可以在临床实践中轻松用于辅助 VRF 检测，因为它不会增加高计算成本，也不会增加患者接受的辐射剂量。
