Medical image analysis in digital histopathology is a currently expanding and exciting field of scientific research. In this work, histopathological image analysis is extensively studied and a systematic framework for computer-based analysis in H&E stained whole slide images of gastric carcinoma is proposed. The exhaustive experimental study comprises of three fundamental stages, namely, preparation of materials, image pre-analysis and analysis of cancer regions. These stages collectively incorporate the understanding, formulation, implementation and evaluation of suitable image analysis tasks required to achieve the defined research objectives in each stage, for example, registration and annotation transformation between whole slide images, cell nuclei segmentation and classification, multiresolution combination of visual information for segmentation enhancement, appearance-based necrosis detection, and cancer classification based on immunohistochemistry. Computerized applications are also demonstrated as an outcome of the conducted research, including computer-aided diagnosis, content-based image retrieval and automatic determination of tissue composition. The research focuses on the development of methods for effective representation and subsequent classification of regions of interest in histopathological image datasets. For this purpose, two image analysis routes, namely, traditional route and deep learning route are investigated. The traditional route consists of handcrafted feature extraction to describe textural, color, intensity, morphological and architectural properties of tissues followed by traditional machine learning methods including support vector machines, AdaBoost ensemble learning and random forests. In this domain, graph-based methods are extensively explored due to their ability to suitably represent the spatial arrangements and neighborhood relationships in tissue regions. A novel graph-based image description method called cell nuclei attributed relational graph is proposed along with multiple variants, for knowledge description of individual component characteristics, spatial interactions and underlying tissue architecture in histopathological images. In the deep learning route, convolutional neural networks are thoroughly investigated. A self-designed convolutional neural network architecture is introduced and analyzed for cancer classification and necrosis detection, also compared with a widely-known framework and ensemble of deep networks. During the detailed investigation, system performance is rigorously analyzed using different experimental aspects, for instance, algorithm parameters, feature configurations and classification strategies. The entire proposed framework is quantitatively evaluated at each stage using a set of performance metrics. A reasonable performance is achieved using the described methods, comparing favorably, and even outperforming the state-of-the-art techniques in certain occasions. The empirical observations, corresponding conclusions, scientific implications and future directions of the research are thoroughly discussed.Collectively, the discussed histopathological image analysis methods in H&E stained gastric cancer whole slide images aim to reduce manual preparation efforts, inspection times, and inter-and intra-observer variability. Moreover, the developed methods can potentially improve the current state of technology through applications such as automatic classification, content-based image retrieval, archiving, bio-banking, marker quantification, detecting malignant changes over time, providing second opinions to pathologists, thereby contributing towards diagnosis, prognosis, education and research in biology and medicine.
@misc{sharma_medical_2017,
title = {Medical image analysis of gastric cancer in digital histopathology: methods, applications and challenges},
author = {Sharma, Harshita},
copyright = {https://creativecommons.org/licenses/by/4.0/},
doi = {http://dx.doi.org/10.14279/depositonce-5888},
language = {en},
shorttitle = {Medical image analysis of gastric cancer in digital histopathology},
url = {https://depositonce.tu-berlin.de/handle/11303/6332},
urldate = {2020-03-04},
year = {2017}
}