Medizinische Universität Graz Austria/Österreich - Forschungsportal - Medical University of Graz
Gewählte Publikation:
Mayrhofer-Reinhartshuber, M.
Pyramid Dimension of Histological Images of the Heart
Doktoratsstudium der Medizinischen Wissenschaft; Humanmedizin; [ Dissertation ] ; 2015. pp.
[OPEN ACCESS]
FullText
- Autor*innen der Med Uni Graz:
- Betreuer*innen:
- Ahammer Helmut
- Asslaber Martin
- Hofer Ernst
- Altmetrics:
- Abstract:
- Myocardial fibrosis is associated with cardiac arrhythmias, its texture influences their incidence. An automated classification system could act as a supporting system for pathologists and help them diagnosing different fibrotic patterns. For that reason an image analysis method for an automated classification of different types of fibrosis in histological images is highly desirable. Fractal analysis quantifies the morphology in images, hence it is a promising approach for the development of such a system. Since medical, especially histological images often have very high resolutions, it is of high relevance to improve existing fractal methods and to develop new approaches for providing fast and accurate results based on these large amounts of data. For fractal dimension estimation of grey value images, three new pyramidal approaches were developed, namely the Pyramid Triangular Prism Method (PTPM), the Pyramid Gradient Method (PGM), and the Pyramid Differences Method (PDM). By using artificially created images the new methods were compared to five standard, non-pyramidal approaches. All new pyramidal methods yielded reasonable values for the fractal dimension D in rather fast computational times. Best results were obtained with the PDM, having a high agreement with theoretical D values and fastest computational times. Based on the new methods, classification algorithms were developed to automatically assign images to five different types of fibrosis. Tests on different image sets showed that the proposed methods resulted in up to 73% correctly (in accordance with human experts) classified images. Although the accordance between the experts was better (up to 90%), utilizing fractal analysis for classification of myocardial fibrosis in digital pathology is definitely a promising approach, should be considered in subsequent developments and may support pathologists in the future.