Development of Methods, Algorithms and Software for the Analysis of Angiogenic Activity of Thyroid Cancer
Tech Area / Field
- INF-IMA/High-Definition Imaging and Displays/Information and Communications
- BIO-PAT/Pathology/Biotechnology and Life Sciences
3 Approved without Funding
National Academy of Sciences of the Republic of Belarus / Institute of Informatics Problems, Belarus, Minsk
- Minsk City Hospital for Oncology, Belarus, Minsk
- University Medical Center / Image Sciences Institute, The Netherlands, Utrecht\nRuprecht-Karls-Universität Heidelberg / Institute for Molecular Biotechnology, Germany, Heidelberg\nEcole Nationale Supérleure des Télécommunications, France, Paris\nHiroshima University / Graduate School of Engineering, Japan, Hiroshima\nUniversity of Essen / Institute of Pathology and Neuropathologie, Germany, Essen\nUniversität Würzburg / Klinik und Poliklinik fur Nuklearmedizin der Universitat Wurzburg, Germany, Würzburg\nImperial College London / Department of Electrical and Electronic Engineering, UK, London
Project summaryProject goal
The Thyroid Tumor Center of Belarus is one of the largest centers of thyroid surgery in the world. More than 14 thousand patients have had surgery in the center after the Chernobyl accident. About 60% of them were treated for the radiation-induced thyroid cancer (Jacob et al., 2006a; Jacob et al., 2006b; Demidchik et al., 2002). Thus, it is vital for the thyroid cancer patients to develop suitable post-operative treatment strategy based on the analysis of a unique archive of histopathological material accumulated during the last twenty years.
The goal of the project is to develop new methods, algorithms and software solutions for quantification and measurement of angiogenic activity of thyroid cancer for predicting clinical course, facilitating surgical strategy, and specifying post-operative treatment. Thus, the project is aimed to an in-depth analysis of color images acquired from one of the world-largest collection of thyroid specimens to advance knowledge on tumor angiogenesis and to develop new treatment modes.
Introduction and Overview
What is angiogenesis. Angiogenesis is a complex multistep process characterized by the formation of new capillaries from the preexisting vascular network. The growth of blood vessels in cancer tumors is 30–40 times faster compared to the vasculature of normal tissues (see Hobson et al., 1984). Several studies have demonstrated a significant correlation between the marked angiogenesis evidenced by a high microvessel density, metastases, and poor prognosis in several tumor types, including breast, prostate, non-small cell lung carcinomas, cutaneous melanomas, and testicular germ cell tumors. It is expected that angiogenesis research will change the face of medicine in the next decades, with more than 500 million people worldwide predicted to benefit from pro- or anti-angiogenesis treatments (Carmeliet, 2005).
Past works. In early angiogenesis assessment works (e.g., Iwahana et al., 1996; Weidner et al., 1993; Fox et al., 1995; Van Der Laak et al., 1998) the computerized image analysis techniques were used mostly for auxiliary purposes such as improving quality of original digital images, highlighting target structures, counting nuclei and counting the immuno-labeled vessels in the most vascularized areas of neoplasm. Later, it was shown that such measurements often lead to conflicting results. The main objective of recent studies (e.g., Goddard et al., 2002, Choi et al., 2005, and Erovic et al., 2005) was to segment and quantify microvessels in tumor angiogenesis based on color image analysis methods.
The demand. Despite the obvious progress, current achievements in the histological image analysis may not be used here for quantification and measurement of angiogenic activity of thyroid cancer. The main reasons are that (a) the project is dealing with radiation-induced thyroid carcinomas, which are characteristic for post-Chernobyl conditions only, (b) the angiogenesis image processing methods developed so far are not suited for a massive computerized analysis of the large biological archive material accumulated in Belarus, and (c) the use of the few conventional angiogenic features only is not sufficient for their joint statistical analysis with the extensive follow-up data of thyroid cancer of post-Chernobyl patients.
Influence to the progress. Providing the community for a new computerized image analysis and diagnosis tools and covering the existing hole in the knowledge of angiogenesis of radiation-induced tumors.
Project participants and their roles. The project team includes specialists from both information technology and oncology. They came from the leading Belarusian centers, namely United Institute of Informatics Problems (UIIP) and Minsk City Hospital for Oncology (MCHO) respectively. The role of UIIP professionals is mostly on developing databases, image processing methods, software solutions, performing angiogenic image analysis and statistics. The role of MCHO is to set up the research hypotheses, select patients, collecting clinical data, preparing biological probes and creation of histological image database. Other activities such as testing research hypotheses and creating prognostic models are performed jointly.
Expected Results and their Application
General results. The project lies in the field of applied research and development and it will be resulted in new methods, algorithms and software for measuring angiogenic activity of thyroid cancer and specifying post-operative treatment.
Specific results. The execution of the six project tasks is resulted in the following specific outcomes.
Task 1: Retrospective clinical database for patients suffering from thyroid cancer and benign tumors in Belarus.
Task 2: A set of samples containing a few thousands immuno-histochemically processed thyroid biological probes.
Task 3: An exhaustive histological database containing tens of thousands of high-resolution color angiogenic images.
Task 4: Color image analysis software for extracting features of angiogenic activity of thyroid tumors.
Task 5: A collection of multivariate statistical analysis procedures and results of the tumor angiogenesis study.
Task 6: Computerized atlas and image retrieval software for computer-aided diagnosis of thyroid carcinomas.
Project consequences. In scientific area the project will contribute towards the advanced color image analysis methods and new knowledge on radiation-induced tumor angiogenesis. In practical domain it will provide certain improvements on diagnosis, prognosis and treatment of thyroid cancer patients. In the field of commerce the project will introduce new products on the market of medical image analysis software and electronic atlases for computer-aided diagnosis.
Application areas. They include general color image analysis problem, angiogenesis in clinics and medical education.
Dissemination. Since both participating institutions represent leading Belarusian centers in their professional areas, this secures a high level of conducted research and legal, duly organized dissemination of the results in Belarus and worldwide.
Meeting ISTC Goals and Objectives
Meeting ISTC goals and objectives is achieved by:
- Providing former weapons scientists and developers of satellite nuclear engines for opportunities to redirect their talents and experience to a highly humanistic and extremely important problem of radiation-induced cancer;
- Integration of project participants into the international community of cancer research by dissemination of new histological image analysis software and novel results on thyroid tumor angiogenesis;
- Supporting applied research for peaceful purposes of public health and nuclear safety in post-Chernobyl era;
- Contributing to the solution of international engineering problem of developing new efficient equipments for medical diagnosis in oncology, computerized prognosis and planning;
- Reinforcing the transition of applied science sector to market-based economies responsive to civil needs.
Scope of Activities
Project scope. The whole bunch of project activities is naturally subpided into the two groups of actions that can be conditionally referred to as engineering and medical. The medical wing (Hospital for Oncology) is mostly responsible for preparing clinical data and input angiogenic microscope images. The engineering wing (Institute of Informatics Problems) is liable for developing image analysis and statistical assessment software. The final study of relationships between the features characterizing the tumor angiogenic activity and other data is performed jointly.
Project efforts. The total number of project participants is 25 with the total project effort of 7261 person*days. It includes 14 engineering participants with the total effort of 4415 person*days and 11 medical with the efforts of 2846 person*days.
Project specificity. The characteristic point of this project is the very large amount of sample preparation, image acquisition, and image analysis efforts. In particular, it is supposed that there will be about 4000-5000 tumor samples prepared and immuno-histochemically processed. Given that the image acquisition protocol requires to collect 10 images for each sample, there must be about 50 thousands of high-resolution microscope images inpidually acquired, stored, processed and measured to provide reliable statistics for every sub-group of patients and each histological cancer type.
Role of Foreign Collaborators
Involvement of foreign collaborators is based on their genuine interests in both innovative methods of sophisticated image analysis and the unique collection of histopathological material. The role of foreign collaborators is
- to provide high level expertise on developing advanced image analysis methods and interpreting angiogenesis data,
- to validate the intermediate project results and to provide advices on further steps,
- to participate in preparation of joint publications resulted from the project and dissemination of the results,
- to provide technical help and occasional, non-systematic use of lab equipment in cases of high mutual interests.
It is anticipated that the collaborators will submit further proposals based on the results obtained within the project (e.g., it is expected a proposal on angiogenesis and nuclear safety will be submitted to EPSRC for a national British funding).
Technical Approach and Methodology
The central idea of the new image analysis approach to be developed in the project is that it does not suppose traditional steps of isolating, segmenting and counting of blood vessels, cells and other histological objects. Instead, it assumes characterization of “histological image patterns” and establishing their quantitative relationships with the target factors (output quantitative measures) on a representative training set. Once found, these quantitative relationships are used for a “direct” computation/prediction of resultant features using specific histological patterns of digital color images under analysis (as opposed to the existing multi-step incremental approaches). The quantitative relationships will be established with the help of (a) statistical univariate methods including conventional linear regression as well as non-linear regression models with polynomial and spline functions; (b) statistical multivariate regression models; and (c) the most recent algorithms that make use the ability of Support Vector Machines (SVM) and some other clustering methods to quantify regression-like relationships (including multivariate ones) between the groups of measurements.
Two kinds of image pattern characterization approaches will be investigated and their efficiency compared to each other and to the conventional methods: (a) Characterization of color image patterns using spatial color co-occurrence matrices (pixel-level descriptors) and (b) Describing histological image patterns based on spatial occurrence of elementary image segments extracted on a pre-processing stage with the help of conventional algorithms (segment-level descriptors).
The above methodology has been introduced very recently and it originates from fruitful ideas of generalized co-occurrence matrices suggested by project participants and collaborators in (Kovalev, Petrou, 1996) and applied to 2D (Kovalev et al., 1998) and 3D (Kovalev et al., 1997) shape analysis, 2D (Kovalev, Petrou, 1996) and 3D (Kovalev, Petrou, 2000; Kovalev et al., 2001; Kovalev et al., 2003, Kovalev and Kruggel, 2007) gray-scale texture characterization, color image retrieval (Kovalev, Volmer, 1998) and blood cell recognition (Kovalev et al., 1996; Ahn et al., 2001). This new methodology partly correlates with some latest ideas in this line of enquiry suggested by others, for instance, with the method of analysis of relationships between the image patterns of mast cells and endothelial cells around the blood vessels in melanoma samples (Guidolin et al., 2006). The project participants have also an extensive experience in the field of image processing (Tuzikov et al., 2006), feature extraction (Sheynin, Tuzikov, 2003; Goncharenko, Tuzikov, 2006) and applied medical imaging (Sanko, Tuzikov, 2006; Snezhko, Tuzikov, 2006; Bogush, Tuzikov, 2006) directly related to this proposal subject.
The International Science and Technology Center (ISTC) is an intergovernmental organization connecting scientists from Kazakhstan, Armenia, Tajikistan, Kyrgyzstan, and Georgia with their peers and research organizations in the EU, Japan, Republic of Korea, Norway and the United States.
ISTC facilitates international science projects and assists the global scientific and business community to source and engage with CIS and Georgian institutes that develop or possess an excellence of scientific know-how.