Computerized identification method for current dental intraoral radiographs
スポンサーリンク
概要
- 論文の詳細を見る
Dental intraoral radiographs are saved random-ly in a PACS (Picture Archiving and Communication System). Therefore, dental radiologists have to arrange manually the dental intraoral radiographs on the PACS in the interpreta-tion of the dental intraoral radiographs. This task is bother for dental radiologists. The purpose of this study was to de-velop a computerized method for identifying a current image as the corresponding previous image which was taken at ap-proximately the same position in order to arrange automati-cally current intraoral radiographs on the PACS based on the previous images which were arranged in the last interpreta-tion. Our database consisted of 56 current and 56 previous dental intraoral radiographs. In our proposed method, the edges of teeth crowns were first enhanced by applying a Sobel filter in the vertical direction to images. The edges of teeth crowns were then segmented by applying a gray-level thre-sholding technique to the enhanced image. Approximate straight line for tops of the teeth was drawn based on a least squares method with the segmented edges. We calculated the angle between the approximate straight line and the horizontal line and rotated the original dental radiograph by an affine transformation to make the growing direction of teeth vertical. The average pixel values in the vertical direction on each x-coordinate became low between adjacent teeth because teeth regions had higher pixel values than other tissues. Therefore, individual tooth regions were divided by the vertical lines through the x-coordinates with local minimum values on the average pixel values. For all combination of current and pre-vious images for a patient, the correlation coefficients between two images were calculated based on the pixel values after aligning images such that the number of the taken teeth be-comes the same on the comparing regions. The combination of current and previous images with the highest correlation coefficient was identified as the corresponding images taken at approximately the same position. With the proposed method, identification accuracy was 80.4% (45/56). The proposed method was shown to have high identification accuracy, would be useful in the terms of the efficiency of diagnosis in the PACS.
論文 | ランダム
- 長野県小谷村に残る伝統的茅場の植物相
- 動脈硬化の病態における慢性炎症の役割
- リポソームDDSの開発研究 : トランスレーショナルリサーチの実現を目指して
- ニコチン生合成のマスター転写因子
- 名古屋議定書に関する政府間委員会第1回会合