Automated quantitative diagnosis system for dental diseases by optimum image processing

许多研究报告称，数字牙科成像系统的诊断准确率可与胶片相媲美。这可能是由于数字数据的图像显示不充分，本研究的目的是通过应用数字数据的最佳图像显示，从数字牙科成像系统中获取最大的图像信息。进一步的目标是开发一种使用龋齿诊断算法和优化图像的计算机辅助诊断系统。主要研究结果如下：1.数字图像数据采集制作了27颗拔除的牙齿，其中包括不同等级的龋齿。用数字化牙科影像系统对两颗牙进行摄片。曝光时间分8个步骤变化，其中包括最佳曝光。将牙齿切成两半，在体视显微镜下观察近牙面的组织病理学变化。2.近牙面自动轮廓提取算法为了实现近牙面的自动检测，提出了一种基于小波变换的近牙面自动检测算法。3.龋病自动诊断算法在邻面位置设置感兴趣区域(ROI)。提出了一种测量垂直于齿面的像素值变化的算法。通过沿近牙面移动，像素值的轮廓改变。沿近牙面获得了一定厚度的磨削。确定了像素值变化的几个代表性数值。4.对该系统的诊断准确性进行了评估。本系统的诊断准确率与观察者相当，优于某些观察者。我们认为，该系统可以作为一种计算机辅助诊断系统用于近端龋病的诊断。

Many studies have reported that diagnostic accuracy of the digital dental imaging system is comparable to films. It may be due to the inadequate image display of the digital data The purpose of this study was to obtain maximum image information from digital dental imaging system by applying optimum image display of the digital data. Further aim was to develop a computer-assisted diagnosis system using caries diagnosis algorithm and optimized images. The results obtained are as follows.1.Collection of digital image dataTwenty-seven extracted teeth including caries of various grades were prepared. Pair of two teeth were radiographed with digital dental imaging system. Exposure time was varied in 8 steps including optimum exposure. Teeth were cut into halves and histopathology of proximal surfaces was assessed under the stereomicroscope. The final diagnoses were used as the "gold standards".2.Algorithm of automated outline extraction of proximal tooth surfaceTo detect proximal tooth surface automatically, an algorithm was developed using Wavelet transforms. It was possible to extract exact proximal tooth surface automatically by using this method.3.Algorithm of automated caries diagnosisA region of interest (ROI) was set at the interproximal position. An algorithm was developed to measure the pixel value changes vertical to the tooth surface. By moving along the proximal tooth surface, a profile of the pixel value changes. along the proximal tooth surface with certain thickness was obtained. Several representative values of the pixel value change have been determined.4.Evaluation of the diagnostic accuracy of this systemAutomated caries diagnosis was performed using the representative values extracted above. The diagnostic accuracy of the present system was comparable to that of the observers and better over some of the observers. We consider this system may be useful as a computer-assisted diagnosis system for proximal caries diagnosis.