@MASTERSTHESIS\{IMM2006-04817,
    author       = "E. M. Jensen",
    title        = "Automated Quantitative Histomorphometry in Osteoarthritis",
    year         = "2006",
    keywords     = "Osteoarthritis, image analysis, quantitative measurements, histomorphometric, semi automation, articular cartilage, chondrocytes, cellularity, cartilage matrix, bone density.",
    school       = "Informatics and Mathematical Modelling, Technical University of Denmark, {DTU}",
    address      = "Richard Petersens Plads, Building 321, {DK-}2800 Kgs. Lyngby",
    type         = "",
    note         = "Supervised by Bjarne Kj{\ae}r Ersb{\o}ll, {IMM,} {DTU,} and Michael Grunkin, Visiopharm A/S",
    url          = "http://www2.compute.dtu.dk/pubdb/pubs/4817-full.html",
    abstract     = "In the development process of drugs for different diseases the pharmaceutical companies spend a lot of resources in the preclinical phase when researching for compounds as drug candidates. In the matter of osteoarthritis (degenerative joint disease), part of the research includes test on mice which spontaneously develop the disease. It is difficult to determine the level of osteoarthritis (OA) influence in each mouse, and currently this is being done manually. Defining standards for quantitative assessment to replace current manual scoring systems and improve precision, is of great interest. Also semi automation of the assessment is demanded to speed up the process, and to eliminate bias caused by the subjective way of manual evaluation.

In this thesis, a set of quantitative histomorphometric features have been analyzed and validated, for use as {OA} end points. A dataset of images acquired by microscope of Hematoxilin/Eosin-stained histological sections from the left knee joint of {OA} affected mice, have been used for this work. Features have been extracted by measurements from the images using software implemented image analysis methods. The features has been validated as end points by statistics using manually given pathology scores.

Several biological parameters in the articular cartilage in the joint are known to be connected to the presence of {OA,} of which the following have been evaluated: The condition of the cartilage matrix structure, cellularity of chondrocytes inside the cartilage and the structure of the subchondral bone. The evaluation was focused on the medial tibia part of the knee joint.

A general measurement model has been proposed and used for all the measurements by a base line and a left offset found by modified Cusum control chart. The model takes advantage of advanced segmentation of the tissue in the images. Also the advantage of dedicated software algorithms and high processing power has been used to analyze features with possible correlation to the {OA} impact.

The measurements of the cartilage matrix structure were done by examining irregularities in the cartilage surface. A strong irregularity feature has been defined with a significant correlation to the manually given cartilage pathology score. The cellularity was measured as a fraction of chondrocyte area and cartilage area. From this measurement a feature based upon the slope in chondrocyte density was found, which yielded a significant correlation with the manual given cellularity pathology score. The bone density in the subchondral bone was measured using the model and a high correlation with the manually determined bone pathology score was found.

The found end points have been further validated by determining the treatment effect of a used drug compound. All end points indicated significant treatment effect. A throughout age dependency study has been performed using the end points for assessment. The results indicated significant correlation between known {OA} age related trends and the end points.

A prototype software module for Visiopharm Integrator System was designed and developed using standard methods of Object Oriented Analysis and Design. The module implements the described measurement methods and end points. A semi-automatic solution embeds the module, and offers a strong tool for preclinical research of osteoarthritis."
}