Automatic identification of clue cells in microscopic leucorrhea images provides important information for evaluating gynecological diseases. Traditional manual microscopic examination of Gram-stained vaginal smears is adopted by most hospitals for identifying clue cells; however, it is both complex and time-consuming. In order to solve these problems, an automatic identification of clue cells in microscopic leucorrhea images based on machine learning is proposed in this paper. First, the Otsu threshold method is used to segment regions of interest (ROI) in image preprocessing according to the morphological features of clue cells. Then, Gabor, HOG and GLCM texture features are extracted to describe irregular edges and rough surfaces of clue cells. Finally, a SVM classifier using a hybrid kernel function by linearly weighted RBF and polynomial kernels is trained to identify clue cells rapidly and conveniently. In experiments, the method using GLCM texture features and a hybrid kernel function of SVM achieved 94.64% accuracy and 94.92% recall rate, which was better than methods using Gabor or HOG texture features and a single kernel function of SVM.
Fecal microscopic examination is a routine examination item to determine whether the digestive system is normal by analyzing formed elements. Traditional method is that doctor uses microscope eyepiece to observe sample smears. The efficiency is low, and examination results depend on doctor's experience level. Therefore, intelligent identification of formed elements is the main development direction of current fully automated fecal instruments. Unlike blood or urine samples, human fecal samples contain a lot of impurities, and sample stratification phenomenon is serious. So image quality assessment methods are difficult to find the sharpest image, affecting effectiveness of intelligent identification algorithm. In this paper, the microscopic image autofocus technology for human fecal samples is studied and divided into two parts: location and photographing. In location process, we use SMD algorithm to determine sample photographing interval. In photographing process, microscope platform zigzagged move in the interval to obtain each view's successively image sequences of different focal lengths. In order to accurately find the sharpest image in image sequence, we compared the difference between human eyes with 31 types of no-reference image quality assessment methods based on entropy, gradient, color, edge, contrast, similarity, and transform domain. Finally an improved Local TV algorithm was chose. Experimental results show that the improved Local TV algorithm is insensitive to changes in sample concentration with good robustness, and the accuracy rate can reach 94.26%. Our experimental results have some reference value for other focusing problems of complex microscopic images.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.