Automated diagnosis system for dental diseases using functions of the humanvisual system

本项目的目的是从普通观察者那里获得人类视觉系统的功能，并将其作为半自动化系统应用于近端龋齿诊断。人类视觉系统对近端龋齿诊断的影响本实验采用模拟胶片和数字成像胶片两种记录介质。材料包括27颗从青少年中拔出的第三磨牙。利用这些设备和材料，共获得了594张数字图像和189张胶片照片。7名口腔放射科医生一次评估一个系列。在记录系统每次曝光时获得受试者工作特征（ROC）曲线。在本实验中，我们得出结论，记录介质所记录的辐射对比度决定了龋齿诊断所需的信息。人类视觉系统对龋齿诊断的影响最小。基于以上实验结果，我们开发了一套半自动的近端龋齿诊断系统。在该系统中，龋的诊断算法为：1)近端表面轮廓的提取2)沿水平轴的灰度测量3)沿近端表面的灰度轮廓4)对轮廓的多项式曲线拟合5)从函数中提取参数6)利用上述参数对龋的判别函数。特别是在采用自动曝光补偿程序的系统中效果最好。结论该半自动化系统可用于近端龋的诊断，特别是具有自动暴露补偿功能的系统。

The purpose of this project was to obtain the function of the human visual system from the average observer and apply it to the proximal caries diagnosis as a semi-automated system.1.Effects of the human visual system on the proximal caries diagnosisTwo types of recording media-the analogue film and the digital imaging films- were used for this experiment. The materials consisted of 27 third molar teeth extracted from young adolescents.Using these equipments and materials, a total of 594 digital images and 189 film radiographs were obtained. Seven oral radiologists evaluated one series at one time. Receiver operating characteristic (ROC) curves were obtained at each exposure of the recording system.In this experiment, it was concluded that radiation contrast registered by the recording medium determines the information required for the caries diagnosis. The effect of human visual system on the caries diagnosis was minimum.2.Semi-automated system for the proximal caries diagnosisUsing the results from the above experiment, we have developed the semi-automated system for the proximal caries diagnosis.In this system, the algorithm for the caries diagnosis is following :1) extraction of the contour of the proximal surface2) gray level measurement along the horizontal axis3) gray level profile along the proximal surface4) polynomial curve fitting to the profile5) extraction of the parameters from the function6) discrimination function for the caries using the above parametersDiagnostic accuracy obtained from the semi-automated system was comparable to that of the observers. Especially, it was most effective in the system using automated exposure compensation program.ConclusionThe present semi-automated system may be useful for the proximal caries diagnosis, especially in the system with automated exposure compensation.