Leucorrhea-wet-film recognition based on coarse-to-fine CNN-SVM

Candida and leukocyte are two important indicators in the diagnosis of gynecological inflammation in microscopic images using the leucorrhea wet film. However, in the microscopic image of leucorrhea wet films, insignificant contrast between target and background, slight differences in texture, weak edges, drab gray on the whole, etc., make intelligent detection of white blood cells and Candida in the microscopic image of leucorrhea wet film extremely difficult. To tackle the problem, we propose a detection method based on coarse-to-fine CNN-SVM, in which the films are pre-filtered with a morphological opening operator, and then white blood cells are identified by using Hough circle detection, and finally, the feature extraction and classification of Candida are implemented based on coarse-to-fine CNN-SVM. Experminents results are also provide to demonstrate the performance of the proposed method.