Automatic Quantification of Epidermis Curvature in H&E Stained Microscopic Skin Image of Mice

Hussein, Saif and Jassim, Sabah A. and Al-Assam, Hisham (2017) Automatic Quantification of Epidermis Curvature in H&E Stained Microscopic Skin Image of Mice. In: Medical Image Understanding and Analysis. MIUA 2017. Communications in Computer and Information Science, 723 . Springer International Publishing, Cham, pp. 935-945. ISBN 978-3-319-60964-5

Preview

Text Hussein_2017_Automatic.pdf Download (1MB) | Preview

Abstract

Changes in the curvature of the epidermis layer is often associated with many skin disorders, such as ichthyoses and generic effects of ageing. Therefore, methods to quantify changes in the curvature are of a scientific and clinical interest. Manual methods to determine curvature are both laborious and intractable to large scale investigations. This paper proposes an automatic algorithm to quantify curvature of microscope images of H&E-stained murine skin. The algorithm can be divided into three key stages. First, skin layers segmentation based on colour deconvolution to separate the original image into three channels of different representations to facilitate segmenting the image into multiple layers, namely epidermis, dermis and subcutaneous layers. The algorithm then further segments the epidermis layer into cornified and basal sub-layers. Secondly, it quantifies the curvature of the epidermis layer by measuring the difference between the epidermis edge and a straight line (theoretical reference line) connecting the two far sides of the epidermis edge. Finally, the curvature measurements extracted from a large number of images of mutant mice are used to identify a list of genes responsible for changes in the epidermis curvature. A dataset of 5714 H&E microscopic images of mutant and wild type mice were used to evaluate the effectiveness of the algorithm.

Actions (login required)

View Item