Skip to main content
Open Access Publications from the University of California

UC Irvine

UC Irvine Electronic Theses and Dissertations bannerUC Irvine

Analysis of Near-Infrared Image to Diagnose Maxillary Sinusitis


Introduction: Sinusitis is one of the most common chronic illnesses. Computed tomography is the most common imaging method in sinusitis detection. Objectives: The main purpose of this study is to determine the diagnostic usage of NIR imaging in maxillary sinusitis. It is also determined which feature set produces the most efficient classification results. In addition, it is investigated whether the color normalization of the NIR images affects on the results. Results: Histogram mean, texture inertia, and texture entropy are the most efficient features in data discrimination. The best sensitivity in maxillary sinusitis detection is 76% produced by using asymmetry indicator values of histogram mean feature extracted from the original images. In addition, the discrimination functionality of the selected feature set is degraded by color normalization. Methods: After the NIR images are prepared, their regions of interest (ROI) are selected manually. Then several features are extracted from the images. The values are used to measure a feature-based asymmetry indicator according to the left and right maxillary sinuses in each image. Also, the images of test class (test set) can be classified by having the range of the asymmetry indicator for control and sever images (train set). The classification correctness metrics are calculated to evaluate the diagnostic role of NIR images in sinusitis disease. Conclusion: It is possible to detect sinusitis using NIR imaging with the sensitivity of 76%. The most effective feature for maxillary sinusitis detection is histogram mean feature. Color normalization is not recommended to be applied on the images.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View