Does Two-Class Training Extract Real Features? A COVID-19 Case Study

Abstract

Diagnosis aid systems that use image analysis are currently very useful due to the large workload of health professionals involved in making diagnoses. In recent years, Convolutional Neural Networks (CNNs) have been used to help in these tasks. For this reason, multiple studies that analyze the detection precision for several diseases have been developed. However, many of these works distinguish between only two classes: healthy and with a specific disease. Based on this premise, in this work, we try to answer the questions: When training an image classification system with only two classes (healthy and sick), does this system extract the specific features of this disease, or does it only obtain the features that differentiate it from a healthy patient? Trying to answer these questions, we analyze the particular case of COVID-19 detection. Many works that classify this disease using X-ray images have been published; some of them use two classes (with and without COVID-19), while others include more classes (pneumonia, SARS, influenza, etc.). In this work, we carry out several classification studies with two classes, using test images that do not belong to those classes, in order to try to answer the previous questions. The first studies indicate problems in these two-class systems when using a third class as a test, being classified inconsistently. Deeper studies show that deep learning systems trained with two classes do not correctly extract the characteristics of pathologies, but rather differentiate the classes based on the physical characteristics of the images. After the discussion, we conclude that these two-class trained deep learning systems are not valid if there are other diseases that cause similar symptoms.

Reliability Evaluation of the Factors That Influenced COVID-19 Patients’ Condition

“…Therefore, the structure function building can be interpreted as a classification task for incomplete or uncertain data. It is a typical task of Data Mining [37,44,45]. One of the approaches for solving this task is the application of a Fuzzy Decision Tree [14,44].…”

“…In this paper, a new method for the structure function construction based on uncertain and incomplete data is considered. This method was developed based on knowledge of reliability engineering [26,27,42] and Data Mining [37,44,45]. The proposed method, from a point of view of reliability analysis, was elaborated taking into account the problem of the system mathematical model representation based on uncertain data.…”

“…In this paper, we propose a new method for the structure function construction for the mathematical representation of a system based on incompletely specified and uncertain data, which can be used for different types of systems in healthcare. For example, uncertainty and incompleteness are typical for data related to coronavirus disease problems [10,11,37,38]. The important aspect of the proposed method is interpretation of the structure function as classifier, which divides all possible system states into several predefined classes (for example, fault states and working states) [25,33].…”

“…This problem is typical for Data Mining. There are many Data Mining-based methods that allow classifying of uncertain and incompletely specified data [37,39]. In this paper, the Fuzzy Decision Tree is used for the classification of initial data to construct the structure function.…”

“…In this paper, the proposed method for the construction of the mathematical representation of a system in the form of the structure function, based on uncertain and incompletely specified data, is used for evaluating COVID-19 patients' condition. The data about coronavirus disease is uncertain and incompletely specified at the presented time [10,22,37]. The initial data for some investigation of the problems correlated with COVID-19 were collected based on questionnaires [9] or expert knowledge [22].…”