![]() The colorectal-cancer (CRC) histology data used in this study were originally studied in 16. Three public databases are used in this study and descibed as follows. Comparative results obtained from testing three public datasets of histopathological images, including haematoxylin-and-eosin (H&E) stained tissue images of colorectal cancer, H&E stained tissue images of heart failure, and immunohistochemistry (IHC) stained tissue images of rectal cancer, show the capability of achieving very accurate classification by the current approach. The class modeling is then designed for training the networks. The networks receive input as the combination of multiple features extracted in time-frequency and time-space domains of the time series transformed from the images. ![]() In this study, recurrent neural networks that learn time-frequency and time-space features for classification of time series or sequential data 15 is further developed for classifying histopathological images. More recently, fusion of deep-learning features was performed for classifying histopathological images of breast tissue 13 and texture features extracted from histopathological tissue images of prostate cancer were used for image classification with support vector machines to provide Gleason scores to the patients’ whole slide images, and the results found to be better than the use of deep learning 14. Some of these works include a self-designed convolutional neural network (CNN) model for necrosis detection in whole-slide images of gastric cancer 8, the use of a CNN model for pathology-based prediction of survival outcome of patients with lung cancer 9, a pre-trained CNN (Inception v3) for detecting cancer subtype or gene mutations from histopathological images of non-small cell lung cancer 10, pre-trained CNNs for identifying histologic growth patterns of lung cancer 11, and Bayesian CNN for classifying histopathological images of colorectal cancer 12. A type of advanced machine-learning method such support vector machines (SVM) was utilized to develop a system for classifying normal tissue and tissue lesions from liver, lung, spleen, and kidney of bovine animals into different histologic categories 4.Īs image processing and classification using deep learning has been realized as a major direction of research in medical prognostics and health management 5, using the state-of-the-art methods in artificial intelligence (AI) for pattern classification, several deep-learning models have recently been used for classification in digital pathology 6, 7. Not only from the perspective of the ability for rendering accurate diagnosis, but the automated analysis can also provide insights into disease mechanisms for understanding biological abnormalities, optimal clinical patient-specific treatment, and biomarker discovery.Īutomated image analysis of spatial structures of histopathological images were carried out in works reported in 2, 3. The benefits of automated image analysis of histopathological images are multi-fold 1. Histology is the study of the microscopic anatomy of biological tissues, while histopathology is a field of histology that involves the study of diseased tissue. These factors have arisen the need for automated quantification of digital pathology data. Conventional pathological quantification, which is based on the expertise of pathologists, is subjective in assessment, time-consuming for the analysis of large volumes of data, and may encounter difficulties when the reproducibility of results is desired. Image analysis in pathology is an important task that helps provide pathologists with quantitative information to be discovered in complex characteristics of pathology images. The proposed approach has the potential to be an AI tool for robust classification of histopathological images. Tests on several datasets of histopathological images of haematoxylin-and-eosin and immunohistochemistry stains demonstrate the strong capability of the artificial intelligence (AI)-based approach for producing very accurate classification results. ![]() Furthermore, unlike conventional classification practice, a strategy for class modeling is designed to leverage the learning power of the TF-TS LSTM. The deep learning is empowered by the use of sequential time-frequency and time-space features extracted from the images. Here, the time-frequency time-space long short-term memory network (TF-TS LSTM) developed for classification of time series is applied for classifying histopathological images. Particularly for cancer research, precise classification of histopathological images is the ultimate objective of the image analysis. Image analysis in histopathology provides insights into the microscopic examination of tissue for disease diagnosis, prognosis, and biomarker discovery.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |