Few-Shot Learning based on Residual Neural Networks for X-ray Image Classification

Currently, deep learning is widely used in the field of medicine, which in turn includes radiology. This paper considers the problem of the classification of X-ray images and the lack of images of specific classes. The classes included COVID-19 and Normal X-ray scans. To solve the problems, we propose few-shot learning that is based on different Residual Convolutional Neural Network models with different complexities. The method is designed for the datasets that have small amount of samples of a specific class and a larger amount of instances of another class. The utilization of few-shot learning can solve the issues of the balance of X-ray datasets. The Residual Convolutional Neural Network models we used are as follows: ResNet-50, ResNet-101, and ResNet-152. The architectures had been used to extract the features from the images that were used later. The latter model has the highest complexity, while the former has the lowest complexity, respectively. The obtained results include the highest accuracy of 97.7% for 10 shots of COVID-19 positive X-ray images. The accuracy was achieved using ResNet-101 model. The highest result for ResNet-152 model was 95.6 %. However, on average, the model achieved the highest accuracy. ResNet-50 model provided the least accurate results, however, it is less complex which provides faster performance. One can also notice that with the higher number of COVID-19 positive shots that were used for training, the accuracy also gets higher. To provide transparency to our solution, we furthermore created t-distributed stochastic neighbor embedding visualization. This showed us that the system could separate the two classes into two distinct clusters. Overall, the results imply the efficiency of the solution that was proposed in the study. © 2022 IEEE.

Few-Shot Learning Approach for COVID-19 Detection from X-Ray Images

The end of 2019 and the beginning of 2020 were accompanied by an exponential spread of COVID-19 infection (a viral disease). This later led to a pandemic situation all over the planet. Such a rapid infection of people with the virus (SARS-CoV-2) from each other was caused by the fact that the symptoms of this disease are very similar to ordinary ARVI (acute respiratory viral infection). This in turn complicates the identification of a patient with a new virus. In order to isolate and contain the further spread of the virus, effective and rapid methods are needed to identify patients at an early stage. In our research work, we propose to use the few-shot method. This method is effective with a small amount of input data, training with few-shot is aimed at creating accurate machine learning models with less training data. Since the size of the input data is a factor determining the cost of resources (such as time costs), it is possible to reduce the cost of data analysis by using few-shot learning. The obtained results include the highest accuracy of 97.7% for 10 shots of COVID-19 X-ray images, which implies the effectiveness of the proposed approach. Notably, it was discovered that the accuracy of the approach directly correlates with the number of COVID-19 samples used for training. © 2021 IEEE.

COVID-19 classification based on CNNs models in CT image datasets

The COVID-19 coronavirus pandemic was a global challenge to the whole society and at the same time created a unique situation for the development of science, scientific communication and open access to scientific information. At the beginning of 2019 the world has faced a pandemic of Covid-19 coronavirus. The coronavirus impacted dramatically lives of majority people around the globe. Deep learning methods allow automatic classification of the coronavirus disease from the computer tomography (CT) scans of the lung. In our work we test several popular convolutional neural network (CNN) models to classify slices of the CT scans. In this study we indicate that the VGG-19 model gives the best classification accuracy among the other tested models such as DenseNet201, MobileNetV2, Xception, VGG-16 and ResNet50v2. In particular, the model achieves the accuracy of 99.08% for CovidX CT Dataset and 98.44% for SARS-CoV-2 CT dataset and 92.30% for UCSD COVID-CT dataset. Additionally, our results include 3D heatmaps that explain classification results for each individual model, showing regions of the lung affected by the coronavirus. © 2021 IEEE.