Hyphae Detection in Fungal Keratitis Images With Adaptive Robust Binary Pattern

Fungal keratitis is an inflammation of the cornea that results from infection by fungal organisms. It has a high rate of blindness, which makes the accurate diagnosis of fungal keratitis important. Confocal microscopy is an optical imaging technique that assists doctors in diagnosing fungal keratitis, and cornea images obtained by confocal microscopy can be used to detect hyphae. The current challenges are how to classify normal cornea images with nerves and abnormal cornea images with hyphae and how to detect the hyphae in a complicated background. To address this problem, this paper proposes a novel automatic hyphae detection method that assists doctors in making diagnoses. It includes two primary steps: texture classification of images and hyphae detection. In texture classification step, first, after image enhancement using a subregional contrast stretching algorithm, an adaptive robust binary pattern (ARBP), which is an improvement on the adaptive median binary pattern (AMBP), is proposed and adopted to extract texture features; and a support vector machine model is used to classify the normal and abnormal images. In the hyphae detection step, binarization and a connected domain process are used to further enhance the targets, and a line segment detector algorithm is adopted to detect the hyphae; then, the hyphal density is defined to quantitatively evaluate the infection severity. The contributions of this study include the improvement of the AMBP and the design of a novel framework. ARBP can extract effective texture features of images with relatively bright and small targets. The experimental results demonstrate the effectiveness of the proposed framework.