• 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • a School of Computer Science University


    a School of Computer Science, University of Nottingham, Nottingham, UK b Mathematics Department, Faculty of Science, University of Assiut, Assiut, Egypt c Laboratory of Image & Data Analysis, Ilixa Ltd., Nottingham, UK d Faculty of Medicine, M.K. Ciurlionio 21, LT-03101 Vilnius University, Vilnius, Lithuania e National Center of Pathology, a liate of Vilnius University Hospital Santaros Clinics, P. Baublio 5, LT-08406 Vilnius, Lithuania
    f Nottingham Molecular Pathology Node and Academic Unit of Pathology, Division of Cancer and Stem Cells, University of Nottingham, Queen’s Medical Centre, Nottingham, UK
    Article history:
    Colorectal cancer
    Tissue classification
    Machine learning
    Fuzzy pressure force
    Axis of least inertia
    Feature representation
    Neural network 
    Automated segmentation of tumor Bortezomib (PS-341) from histological images is a fundamental aspiration of digital pathology to improve biomarker assessment and tissue diagnosis. Accurate tumour segmenta-tion is an important step in many automated digital image analysis applications to be used in clinical practice. In particular, segmentation of tumour, non-tumour epithelium and stromal tissue compartments on immunohistochemistry images presents a challenge. Many artifacts, such as staining and/or illumi-nation variations, can confound image analysis. In this paper, we propose a cascade-learning approach which can diminish the impact of these artifacts. It consists of (a) a set of novel invariant features that encodes meaningful information about the appearance and shape of the region of interest and (b) a novel level set formulation where contour evolution is driven by a probabilistic model of the appearance of the region (based on fuzzy c-means). The merit of our approach is that it exploits both appearance and shape information and combines them in the tissue classification framework. We evaluate the performance of our approach on the segmentation of tumour epithelium in colorectal cancer. The experimental results show that our approach is robust to staining differences, additive noise, intensity inhomogeneities, and can cope with a limited number of training samples, when compared to the state-of-the-art tumour ep-ithelial segmentation methods.
    1. Introduction
    Histopathology is the microscopic examination of thin sections of potentially diseased tissue, which have been fixed onto glass slides and stained to reveal particular structural or functional de-tail. With the availability of whole slide scanners, digitized im-ages of those glass slides can be obtained, thereby making tissue histopathology amenable to the application of image analysis al-gorithms. By automating repetitive tasks, improving the accuracy
    ∗ Corresponding author at: Nottingham Molecular Pathology Node, Queen Medi-cal Centre, Nottingham NG8 1BB, United Kingdom.
    E-mail addresses: [email protected], [email protected] (M.M. Abdelsamea), [email protected] (A. Pitiot), rutabarbora. [email protected] (R.B. Grineviciute), [email protected] (J. Besusparis), [email protected] (A. Laurinavicius), [email protected] (M. Ilyas).
    and precision of feature measurements (such as cell count or gland area) and decreasing subjectivity, image analysis could help to im-prove tissue analysis for both biomarker assessment and tissue di-agnosis. Furthermore, mining of the digital data could help identify new diagnostic features for a variety of disease categories.
    The demand for precise segmentation tool increases along with accumulating evidence on the diagnostic value that is produced by analysing the spatial distribution of biomarkers in tumour tis-sue. Diagnostic tool Immunoscore®; introduced solid evidence that quantity and spatial distribution of immune cells through the tu-mour is a reliable prognostic and predictive biomarker. Studies fo-cusing on colorectal cancer note, that numerical expression charac-terizing CD3 + and CD8 + cells, calculated using Immunoscore®; is superior to microsatellite stability status and add additional value to existing TNM system. Standardized way of evaluation is required in order to represent complex interaction of immune cells and tumour tissue. Conventional approach using light microscopy and