The paper explored the problem of automatic diagnosis based on immunohistochemical image analysis. The issue of automated diagnosis is a preliminary and advisory statement for a diagnostician. The authors studied breast cancer histological and immunohistochemical images using the following biomarkers progesterone, estrogen, oncoprotein, and a cell proliferation biomarker. The authors developed a breast cancer diagnosis method based on immunohistochemical image analysis. The proposed method consists of algorithms for image preprocessing, segmentation, and the determination of informative indicators (relative area and intensity of cells) and an algorithm for determining the molecular genetic breast cancer subtype. An adaptive algorithm for image preprocessing was developed to improve the quality of the images. It includes median filtering and image brightness equalization techniques. In addition, the authors developed a software module part of the HIAMS software package based on the Java programming language and the OpenCV computer vision library. Four molecular genetic breast cancer subtypes could be identified using this solution: subtype Luminal A, subtype Luminal B, subtype HER2/neu amplified, and basalt-like subtype. The developed algorithm for the quantitative characteristics of the immunohistochemical images showed sufficient accuracy in determining the cancer subtype “Luminal A”. It was experimentally established that the relative area of the nuclei of cells covered with biomarkers of progesterone, estrogen, and oncoprotein was more than 85%. The given approach allows for automating and accelerating the process of diagnosis. Developed algorithms for calculating the quantitative characteristics of cells on immunohistochemical images can increase the accuracy of diagnosis.
We propose an intelligent system for processing histological and cytological images for diagnosing breast cancers. The main components of this system are based on a fuzzy-knowledge base and databases. The fuzzy-knowledge base contains rules for diagnosis, obtained from expert facts. The databases contain images and their quantitative characteristics. The software is implemented as ImageJ plugin.
In the paper the matching algorithms of region images are analyzed. The work also presents their advantages and disadvantages. The comparing images algorithm of regions is developed on the basis of the measured chords. The comparison regions algorithms are used to evaluate segmentation algorithms in the Gromov-Hausdorff metric. The algorithm of metric evaluation is developed as an example of biomedical images segmentation.
Biomedical images analysis systems become popular nowadays. Their use can reduce the expenses for patients' transporting to hospitals as well as allows receiving some medical advice from experts in relevant fields. Among the deficiencies that may affect the use of these systems we can name the imperfection of the user interface. Overload of visual elements workspaces reduce the effectiveness of software systems to work with them and improve the training of users. The hybrid system includes: dictionary of pathological conditions of cancer, a database of numerical features of microobjects, annotated images, the knowledge base of qualitative features microobjects, conclusion subsystems conclusion.
One of the most significant characteristic features of digital histological images is changing of the form of microobjects (cells) of the same class. The requirement to corresponding points is the stability of their position on the contour, i.e. invariance of their position during rotation, translation. The proposed algorithm of determining the corresponding points on the basis of the central moments is better during translation and rotation transformations. For comparing algorithms the evaluation of symmetric difference between matched regions is used.
1 January 2021 in Ukrainian Journal of Information Technology
This article analyzes the pathological conditions of the breast based on the study of cytological images. Cytological images are a separate class of biomedical images and are used in the diagnosis of cancer. For diagnose precancerous and cancerous conditions and treatment tactics, diagnosticians use cytological, histological, and immunohistochemical images. For automating the process of diagnosis in oncology, automated microscopy systems are used. Automated microscopy systems use computer vision algorithms. Recently, machine learning algorithms have been used to classify images. Microscopic image processing is a complex and time-consuming process, as the images are characterized by high noise levels and the absence of clear contours of cell nuclei. To calculate the quantitative characteristics of cell nuclei cytological images, the method for calculating the quantitative characteristics of cell nuclei based on image filtering algorithms and their automatic segmentation has been developed. An U-Net convolutional neural network architecture has been developed for cell nucleus segmentation. In this work, the method of processing cytological images is developed. The method consists of six stages. The first step is to load the image into the computers memory. In the second stage, the images are preprocessed. The third stage is the automatic segmentation of images based on the convolutional neural network of the U-Net type. In the fourth stage, the quantitative characteristics of cell nuclei are calculated. In the fifth stage, the quantitative characteristics of the cell nuclei are stored in a database. In the sixth stage, linear regression algorithms are used to obtain quantitative characteristics of cell nuclei. Currently, linear regression is one of the common approaches of machine learning to data analysis. In this work, the comparative analysis of the quantitative characteristics application in cell nuclei is carried out based on linear regression. The scientific novelty of the work is development the method for calculating the quantitative characteristics of cell nuclei, which includes stages of image filtering and automatic segmentation based on the use of a neural network such as U-Net. The practical significance of the work is the software implementation of the preprocessing modules and linear regression. In particular, investigated that the set of parameters "area, length of the main axis" has 1.4 times less RMSE error compared to the set "area, perimeter".