Advanced automatic method for detecting maxillary sinusitis on dental panoramic radiographs

DOI Open Access
  • MIKI Yuma
    Department of Intelligent Image Information, Graduate School of Medicine, Gifu University
  • HARA Takeshi
    Department of Intelligent Image Information, Graduate School of Medicine, Gifu University
  • MURAMATSU Chisako
    Department of Intelligent Image Information, Graduate School of Medicine, Gifu University
  • HAYASHI Tatsuro
    Media Co., Ltd.
  • KATSUMATA Akitoshi
    Department of Oral Radiology, Asahi University School of Dentistry
  • ZHOU Xiangrong
    Department of Intelligent Image Information, Graduate School of Medicine, Gifu University
  • FUJITA Hiroshi
    Department of Intelligent Image Information, Graduate School of Medicine, Gifu University

Bibliographic Information

Other Title
  • 歯科パノラマX線写真における上顎洞炎の検出手法の高度化

Search this article

Abstract

In this study, we propose an improved method to detect abnormal regions in maxillary sinuses by accurate determination of regions of interest （ROIs） based on the anatomical information. The ROIs are set based on the automated search of a hard palate and the average widths of tooth crowns. Preferably, dental panoramic radiographs are obtained with the Frankfort plane aligned horizontally. In order to manage the ROI setting in images with inappropriate positioning, the ROI locations are individually adjusted on the basis of the relationship between the reference line and the alveolar line. The proposed method was evaluated with two databases, DB‐1 and DB‐2, consisting of 59 and 39 images. Using the proposed method, the average concordance rate （Jaccard Index） was improved by more than 10%. The areas under the ROC curves for the sinusitis detection were comparable （0.86 and 0.82 for the two databases） with the previous method. As a result, it is expected that the reliability of the computer output was increased with the improved ROI determination.

Journal

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top