Elsevier

Information Sciences

Volume 308, 1 July 2015, Pages 125-139
Information Sciences

A software tool for the automatic detection and quantification of fibrotic tissues in microscopy images

https://doi.org/10.1016/j.ins.2014.10.028Get rights and content

Abstract

The high volume of pathological microscopy images deforming tissues make the fast quantification and detection of the corrupted regions extremely difficult. To tackle this problem, we present in this paper an automated computer based tool that allows the easy selection of training regions for the various type of pathologies and adopts dimensionality reduction and classification methods for detecting and quantifying the infected areas. The output of the proposed tool is a classification result superimposed on original images, along with an overall index indicating the severity of the pathology. The experimental results are promising, since the tool exhibits high classification accuracy and the calculated index is compatible with expert physician’s estimation of the degree of pathology.

Introduction

Recent technological developments, enabled the design and implementation of better imaging devices [31] and improved fluorescent probes [9]. This has strongly influenced medical and biological research due to expanding capabilities of the light microscopes. Furthermore, the rapid capacity increase and price decrease of digital storage, enables the digitization and storage of old film-based microscopy images [13]. However, usually the microscopy images that are stored in such large databases they are not annotated or labeled, so they cannot be used in biomedical research.

Another issue is that microscopy imaging experiments have become more complex, resulting in even bigger image data files [10]. Thus, there is a growing need for automated methodologies and tools that can effectively and efficiently characterize high volume microscopy images, in which the pathology and the tissue type may vary [17].

In this work, we focus on developing such a generic tool for the detection and quantification of abnormal regions in microscopy images of different types. In this context, we have tested different classifiers to find the most promising ones and we have used dimensionality reduction techniques that allow as to get accurate results, while at the same time reduce the computational cost.

More specifically, the developed tool allows the user to load a test image or a batch of images, use the Graphical User Interface (GUI) to annotate as normal or abnormal regions in some test images (training set) and automatically train classifiers using appropriate parameters for the algorithms. Upon the completion of the training phase, the user can load all the remaining images to be characterized and the classification result is superimposed on each microscopy image.

In order to evaluate the efficiency of the proposed tool, microscopy images from fibrotic tissues are used as case study in this work. Fibrosis is a process that constructs a fibrous, connective tissue, different from the normal one, over the tissue or the organ by a reactive or reparative process. In this study, we consider two datasets: the first dataset contains Idiopathic Pulmonary Fibrosis (IPF) images, while the second one corresponds to images with Obstructive Nephropathy.

The rest of the paper is structured as follows: In Section 2, we present the related work, while in Section 3, background material regarding the classification and dimension reduction methods is provided. In Section 4, the methodology and workflow of the proposed tool is presented and in Section 5, we present the experimental results for the aforementioned datasets. Finally, Section 6 contains concluding remarks and pointers for future work.

Section snippets

Related work

A collection of generic tools and frameworks for image acquisition, processing and analysis is available to developers in our days. Typical examples are the ImageJ [2] and the OpenCV [6] libraries, which they provide several image processing operators and functionalities. The ImageJ library offers an abundance of image analysis capabilities that range from standard functionalities such as intensity adjustment to advanced features such as object recognition. On the other hand, the OpenCV library

Background information

In this section, we provide a brief description of the classification algorithms and the dimensionality reduction techniques utilized in the proposed methodology.

Proposed methodology

The purpose of this work is the development of a robust tool for quantification and detection of pathologies (abnormalities) that appear in microscopy images of different types of tissue. The basic idea is that the proposed tool will identify and characterize subregions of images as pathological (abnormal) or normal. For this reason, the test images are subdivided into square blocks of the same size and they are annotated by an expert, using the GUI of the tool. This information is utilized for

Datasets and experiments

In this section, the image datasets used in this study are presented. Subsequently, we report extensive experimental results in an attempt to determine the most appropriate combination of classification algorithm and dimension reduction technique for the specific type of images.

Discussion and conclusions

This work is devoted to the development of a computer based tool, which aims at the accurate quantification and detection of tissue pathology in microscopy images. The presented methodology is based on the use of dimensionality reduction techniques and state of the art classification methods. Our main task is to provide a simple Graphical User Interface, for the physicians to use and interact with. The capability for the provision of feedback by the user is a critical feature of this tool.

The

Acknowledgements

The authors would like to thank the European Union (European Social Fund ESF) and Greek national funds through the Operational Program “Education and Lifelong Learning” of the National Strategic Reference Framework (NSRF) – Research Funding Program: Thalis: “Interdisciplinary Research in Affective Computing for Biological Activity Recognition in Assistive Environments”, and the Academy of Finland under The Finnish Centre of Excellence in Computational Inference Research (COIN), for financially

References (40)

  • C.E. Bonferroni

    Teoria statistica delle classi e calcolo delle probabilità

    Pubblicazioni del R Istituto Superiore di Scienze Economiche e Commerciali di Firenze

    (1936)
  • G. Bradski

    The OpenCV Library

    Dr. Dobb’s J. Softw. Tools

    (2000)
  • G. Bueno et al.

    An automated system for whole microscopic image acquisition and analysis

    Microsc. Res. Tech.

    (2014)
  • M. Chalfie et al.

    Green fluorescent protein as a marker for gene expression

    Science

    (1994)
  • J. Chao et al.

    A software framework for the analysis of complex microscopy image data

    IEEE Trans. Inform. Technol. Biomed.

    (2010)
  • S. Haykin

    Neural Networks: A Comprehensive Foundation

    (1998)
  • P.N. Hedde et al.

    Online image analysis software for photoactivation localization microscopy

    Nat. Methods

    (2009)
  • J. Heuser

    How to convert a traditional electron microscopy laboratory to digital imaging: follow the ‘middle road’

    Traffic

    (2000)
  • G. Izbicki et al.

    Time course of bleomycin-induced lung fibrosis

    Int. J. Exp. Pathol.

    (2002)
  • D. Jackson

    Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches

    Ecology

    (1993)
  • View full text