oalogo2  

AUTHOR(S): 

Ryszard Gomolka, Wlodzimierz Klonowski, Anna Korzynska, Krzysztof Siemion

 

TITLE

Automatic Segmentation of Morphological Structures in Microscopic Images of Lymphatic Nodes - Preliminary Results

pdf PDF

KEYWORDS

pixels intensity distribution, RGB space, segmentation, lymph nodes, biopsy

ABSTRACT

Evaluation of microscopic images of lymphatic nodes is time consuming and is usually based on assessment of images stained with multiple types of dyes. Materials and Methods: Herein, we propose a supportive method for segmentation of follicles, medulla and cortex in H+DAB-stained microscopic images of lymphatic nodes. The method is based on sampling of pixels intensities distribution in the RGB space of the original microscopic image and calculating the feature maps for further segmentation. All calculations are performed in MATLAB and results are compared to regions manually segmented by pathologist. Results: Based on 9 sample images of a lymphatic node, our algorithm achieved in average 0.96, 0.45 and 0.47 accuracy in localization of the follicles, cortex and the medulla, respectively. Conclusions: Our algorithm achieved good performance in segmentation of follicular regions and moderate performance in segmentation of cortex and medulla in the images of lymphatic node. Ongoing works are aimed to improve segmentation of cortical and medullar regions. These regions are difficult to distinguish without hematoxilin- and eosin-stained images. Validation of our algorithm is planned on a larger data set.

Cite this paper

Ryszard Gomolka, Wlodzimierz Klonowski, Anna Korzynska, Krzysztof Siemion. (2017) Automatic Segmentation of Morphological Structures in Microscopic Images of Lymphatic Nodes - Preliminary Results. International Journal of Medical Physiology, 2, 5-8