Showing papers by "Nagaaki Ohyama published in 2019"

Elastic and collagen fibers discriminant analysis using H&E stained hyperspectral images

Abstract: Hematoxylin and eosin (H&E) stain is one of the most common specimen staining methods in pathology diagnosis due to the capability to show the morphological structure of tissue. However, the appearance of the specific component, i.e., elastic fibers might not be recognized easily because have similar color and pattern with ones of collagen fibers. To distinguish these two components, Verhoeff’s Van Gieson (EVG) staining method is commonly used. Nevertheless, procedures of EVG stain are more complex and expensive than H&E stain. In this study, we investigate the possibility to distinguish elastic and collagen fibers from H&E stained images by applying spectral image analysis based on hyperspectral images. With experiments, we measure the transmittance spectral of 61-band H&E stained hyperspectral image, which are converted into absorbance spectral of hematoxylin, eosin, and red blood cell. As many as 3000 sampling pixels both from RGB and hyperspectral images of HE stained specimens were trained using Linear Discriminant Analysis (LDA) to get a discriminant function to classify elastic and collagen components in H&E RGB and H&E hyperspectral images. We conducted verification based on leave-one-out cross-validation of six data sets for evaluation. The verification result both visually and quantitatively compared to EVG stained image shows that the usage of hyperspectral images performs better than RGB images.

Classification of Elastic and Collagen Fibers in H&E Stained Hyperspectral Images

Abstract: This study investigates the classification performance of elastic and collagen fibers using H&E stained hyperspectral images. As many as 1200 sample pixels were trained by using Linear discriminant analysis (LDA) and support vector machine (SVM) methods for reduction and classification. The classification result both using LDA and SVM show that H&E stained hyperspectral images performed better classification than H&E stained RGB image by comparing the classification of EVG stained images visually and quantitatively.