Three-class Classification Task Research Articles

An investigation into augmentation and preprocessing for optimising X-ray classification in limited datasets: a case study on necrotising enterocolitis.

Obtaining large volumes of medical images, required for deep learning development, can be challenging in rare pathologies. Image augmentation and preprocessing offer viable solutions. This work explores the case of necrotising enterocolitis (NEC), a rare but life-threatening condition affecting premature neonates, with challenging radiological diagnosis. We investigate data augmentation and preprocessing techniques and propose two optimised pipelines for developing reliable computer-aided diagnosis models on a limited NEC dataset. We present a NEC dataset of 1090 Abdominal X-rays (AXRs) from 364 patients and investigate the effect of geometric augmentations, colour scheme augmentations and their combination for NEC classification based on the ResNet-50 backbone. We introduce two pipelines based on colour contrast and edge enhancement, to increase the visibility of subtle, difficult-to-identify, critical NEC findings on AXRs and achieve robust accuracy in a challenging three-class NEC classification task. Our results show that geometric augmentations improve performance, with Translation achieving +6.2%, while Flipping and Occlusion decrease performance. Colour augmentations, like Equalisation, yield modest improvements. The proposed Pr-1 and Pr-2 pipelines enhance model accuracy by +2.4% and +1.7%, respectively. Combining Pr-1/Pr-2 with geometric augmentation, we achieve a maximum performance increase of 7.1%, achieving robust NEC classification. Based on an extensive validation of preprocessing and augmentation techniques, our work showcases the previously unreported potential of image preprocessing in AXR classification tasks with limited datasets. Our findings can be extended to other medical tasks for designing reliable classifier models with limited X-ray datasets. Ultimately, we also provide a benchmark for automated NEC detection and classification from AXRs.

Cn2a-capsnet: a capsule network and CNN-attention based method for COVID-19 chest X-ray image diagnosis

Due to its high infectivity, COVID-19 has rapidly spread worldwide, emerging as one of the most severe and urgent diseases faced by the global community in recent years. Currently, deep learning-based diagnostic methods can automatically detect COVID-19 cases from chest X-ray images. However, these methods often rely on large-scale labeled datasets. To address this limitation, we propose a novel neural network model called CN2A-CapsNet, aiming to enhance the automatic diagnosis of COVID-19 in chest X-ray images through efficient feature extraction techniques. Specifically, we combine CNN with an attention mechanism to form the CN2A model, which efficiently mines relevant information from chest X-ray images. Additionally, we incorporate capsule networks to leverage their ability to understand spatial information, ultimately achieving efficient feature extraction. Through validation on a publicly available chest X-ray image dataset, our model achieved a 98.54% accuracy and a 99.01% recall rate in the binary classification task (COVID-19/Normal) on a six-fold cross-validation dataset. In the three-class classification task (COVID-19/Pneumonia/Normal), it attained a 96.71% accuracy and a 98.34% recall rate. Compared to the previous state-of-the-art models, CN2A-CapsNet exhibits notable advantages in diagnosing COVID-19 cases, specifically achieving a high recall rate even with small-scale datasets.

AI-driven material design for tissue engineering a comprehensive approach integrating generative adversarial networks and high-throughput experimentation

This study presents an innovative AI-driven material design approach for tissue engineering, integrating generative adversarial networks (GANs) and high-throughput experimentation (HTE). The research methodology combines synthetic data generation, dimensionality reduction through principal component analysis (PCA), and model evaluation using a random forest classifier. The synthetic data, representative of diverse biomaterial structures, is generated with a three-class classification task. The model undergoes training on PCAtransformed and standardized synthetic data, with evaluation metrics including accuracy, precision, recall, and F1 score. Visualization through scatter plots, confusion matrices, and bar charts provides a comprehensive overview of the proposed approach’s efficacy. Results demonstrate the GAN’s capability to generate diverse synthetic data, the model’s focused learning during training, and its subsequent generalization in the testing phase. Mathematical functions, including sine and cosine, further illustrate fundamental principles, while performance metrics confirm the model’s proficiency in biomaterial classification. This research contributes to the evolving field of AI-driven material design, offering a systematic methodology and visual insights for accelerated and validated biomaterial discovery in tissue engineering applications.

Multi-parametric radiomics of conventional T1 weighted and susceptibility-weighted imaging for differential diagnosis of idiopathic Parkinson’s disease and multiple system atrophy

ObjectivesThis study aims to investigate the potential of radiomics with multiple parameters from conventional T1 weighted imaging (T1WI) and susceptibility weighted imaging (SWI) in distinguishing between idiopathic Parkinson’s disease (PD) and multiple system atrophy (MSA).MethodsA total of 201 participants, including 57 patients with PD, 74 with MSA, and 70 healthy control (HCs) individuals, underwent T1WI and SWI scans. From the 12 subcortical nuclei (e.g. red nucleus, substantia nigra, subthalamic nucleus, putamen, globus pallidus, and caudate nucleus), 2640 radiomic features were extracted from both T1WI and SWI scans. Three classification models - logistic regression (LR), support vector machine (SVM), and light gradient boosting machine (LGBM) - were used to distinguish between MSA and PD, as well as among MSA, PD, and HC. These classifications were based on features extracted from T1WI, SWI, and a combination of T1WI and SWI. Five-fold cross-validation was used to evaluate the performance of the models with metrics such as sensitivity, specificity, accuracy, and area under the receiver operating curve (AUC). During each fold, the ANOVA and least absolute shrinkage and selection operator (LASSO) methods were used to identify the most relevant subset of features for the model training process.ResultsThe LGBM model trained by the features combination of T1WI and SWI exhibited the most outstanding differential performance in both the three-class classification task of MSA vs. PD vs. HC and the binary classification task of MSA vs. PD, with an accuracy of 0.814 and 0.854, and an AUC of 0.904 and 0.881, respectively. The texture-based differences (GLCM) of the SN and the shape-based differences of the GP were highly effective in discriminating between the three classes and two classes, respectively.ConclusionsRadiomic features combining T1WI and SWI can achieve a satisfactory differential diagnosis for PD, MSA, and HC groups, as well as for PD and MSA groups, thus providing a useful tool for clinical decision-making based on routine MRI sequences.

Aflatoxins detection in almonds via fluorescence imaging and deep neural network approach

The detection of aflatoxin B in raw food materials represents a topic of great interest worldwide because of the huge health and economic impact of aflatoxin contamination. In this paper, we present an original approach to aflatoxin detection, using a portable instrument to acquire fluorescence images, among other spectral responses. The acquired images are processed by combining a color space conversion from the RGB scale to Y′CbCr, and a neural network approach to encode a vector of features. After a feature reduction using a Receiving Operating Curve method, two-class and three-class classification tasks of contaminated vs non-contaminated samples are accomplished. This procedure has been applied to artificially contaminated grained almond samples in the range of 0–320.2 ng/g, achieving an overall classification accuracy between 84.7% and 93.0%, depending on the number of classes. Thus, in this setting, we show that good classification performance can be achieved using only an image acquisition and analysis approach. The proposed procedure can represent a cheap, rapid, non-destructive yet sensitive method for the assessment of aflatoxin B contamination in food matrices, and its monitoring and tracing throughout the food chain.

Public Awareness and Sentiment Analysis of COVID-Related Discussions Using BERT-Based Infoveillance

Understanding different aspects of public concerns and sentiments during large health emergencies, such as the COVID-19 pandemic, is essential for public health agencies to develop effective communication strategies, deliver up-to-date and accurate health information, and mitigate potential impacts of emerging misinformation. Current infoveillance systems generally focus on discussion intensity (i.e., number of relevant posts) as an approximation of public awareness, while largely ignoring the rich and diverse information in texts with granular information of varying public concerns and sentiments. In this study, we address this grand challenge by developing a novel natural language processing (NLP) infoveillance workflow based on bidirectional encoder representation from transformers (BERT). We first used a smaller COVID-19 tweet sample to develop a content classification and sentiment analysis model using COVID-Twitter-BERT. The classification accuracy was between 0.77 and 0.88 across the five identified topics. In the sentiment analysis with a three-class classification task (positive/negative/neutral), BERT achieved decent accuracy, 0.7. We then applied the content topic and sentiment classifiers to a much larger dataset with more than 4 million tweets in a 15-month period. We specifically analyzed non-pharmaceutical intervention (NPI) and social issue content topics. There were significant differences in terms of public awareness and sentiment towards the overall COVID-19, NPI, and social issue content topics across time and space. In addition, key events were also identified to associate with abrupt sentiment changes towards NPIs and social issues. This novel NLP-based AI workflow can be readily adopted for real-time granular content topic and sentiment infoveillance beyond the health context.

A Neural Network for Automated Image Quality Assessment of Optic Disc Photographs.

This study describes the development of a convolutional neural network (CNN) for automated assessment of optic disc photograph quality. Using a code-free deep learning platform, a total of 2377 optic disc photographs were used to develop a deep CNN capable of determining optic disc photograph quality. Of these, 1002 were good-quality images, 609 were acceptable-quality, and 766 were poor-quality images. The dataset was split 80/10/10 into training, validation, and test sets and balanced for quality. A ternary classification model (good, acceptable, and poor quality) and a binary model (usable, unusable) were developed. In the ternary classification system, the model had an overall accuracy of 91% and an AUC of 0.98. The model had higher predictive accuracy for images of good (93%) and poor quality (96%) than for images of acceptable quality (91%). The binary model performed with an overall accuracy of 98% and an AUC of 0.99. When validated on 292 images not included in the original training/validation/test dataset, the model's accuracy was 85% on the three-class classification task and 97% on the binary classification task. The proposed system for automated image-quality assessment for optic disc photographs achieves high accuracy in both ternary and binary classification systems, and highlights the success achievable with a code-free platform. There is wide clinical and research potential for such a model, with potential applications ranging from integration into fundus camera software to provide immediate feedback to ophthalmic photographers, to prescreening large databases before their use in research.

D2AFNet: A dual-domain attention cascade network for accurate and interpretable atrial fibrillation detection

Atrial fibrillation is one of the common and potentially dangerous persistent cardiac arrhythmias that are generally associated with the risk of stroke and heart failure. Manual electrocardiography diagnosis is the gold standard for the clinical detection of atrial fibrillation, but it has some drawbacks, such as being time-consuming and prone to misclassification due to inter-patient variability. Due to the powerful ability of deep learning to learn and extract rich features from huge datasets, end-to-end deep learning models are generally designed to detect abnormal atrial fibrillation signals automatically. However, these approaches usually ignore the key factors that feature maps from different channels and sequences may contribute differently to atrial fibrillation detection, making it challenging to implement accurate and interpretable models with better generalization performance. To tackle this challenge, we develop a dual-domain attention cascade D2AFNet for accurate and interpretable atrial fibrillation detection by cascading attention-based bidirectional gated recurrent units and densely connected networks embedded with channel-spatial information fusion modules. The D2AFNet can take full advantage of channel-spatial features to enhance the feature representation in the spatial domain, and then combine with the time series features in the temporal domain to form spatial–temporal fusion attention mechanisms to mine discriminative atrial fibrillation patterns. Besides, the D2AFNet can profoundly explore the different contributions of different spatial and temporal segments of feature maps for excellent interpretation. The proposed D2AFNet method is performed ten-fold cross-validation on the publicly available CPSC 2018 dataset, and achieves the accuracies of 99.49% and 99.28% in the two-class and three-class classification tasks, outperforming cutting-edge atrial fibrillation detection methods. In addition, the powerful generalization performance and inference efficiency of the D2AFNet method are also proved on another publicly available MIT-BIH dataset. The advantages of high performance and interpretability indicate that the D2AFNet method has huge potential in the computer-aided diagnosis of atrial fibrillation.

Multi-Learner Based Deep Meta-Learning for Few-Shot Medical Image Classification.

Few-shot learning (FSL) is promising in the field of medical image analysis due to high cost of establishing high-quality medical datasets. Many FSL approaches have been proposed in natural image scenes. However, present FSL methods are rarely evaluated on medical images and the FSL technology applicable to medical scenarios need to be further developed. Meta-learning has supplied an optional framework to address the challenging FSL setting. In this paper, we propose a novel multi-learner based FSL method for multiple medical image classification tasks, combining meta-learning with transfer-learning and metric-learning. Our designed model is composed of three learners, including auto-encoder, metric-learner and task-learner. In transfer-learning, all the learners are trained on the base classes. In the ensuing meta-learning, we leverage multiple novel tasks to fine-tune the metric-learner and task-learner in order to fast adapt to unseen tasks. Moreover, to further boost the learning efficiency of our model, we devised real-time data augmentation and dynamic Gaussian disturbance soft label (GDSL) scheme as effective generalization strategies of few-shot classification tasks. We have conducted experiments for three-class few-shot classification tasks on three newly-built challenging medical benchmarks, BLOOD, PATH and CHEST. Extensive comparisons to related works validated that our method achieved top performance both on homogeneous medical datasets and cross-domain datasets.

A Hybrid Deep Fused Learning Approach to Segregate Infectious Diseases

Humankind is facing another deadliest pandemic of all times in history, caused by COVID-19. Apart from this challenging pandemic, World Health Organization (WHO) considers tuberculosis (TB) as a preeminent infectious disease due to its high infection rate. Generally, both TB and COVID-19 severely affect the lungs, thus hardening the job of medical practitioners who can often misidentify these diseases in the current situation. Therefore, the time of need calls for an immediate and meticulous automatic diagnostic tool that can accurately discriminate both diseases. As one of the preliminary smart health systems that examine three clinical states (COVID-19, TB, and normal cases), this study proposes an amalgam of image filtering, data-augmentation technique, transfer learning-based approach, and advanced deep-learning classifiers to effectively segregate these diseases. It first employed a generative adversarial network (GAN) and Crimmins speckle removal filter on X-ray images to overcome the issue of limited data and noise. Each pre-processed image is then converted into red, green, and blue (RGB) and Commission Internationale de l’Elcairage (CIE) color spaces from which deep fused features are formed by extracting relevant features using DenseNet121 and ResNet50. Each feature extractor extracts 1000 most useful features which are then fused and finally fed to two variants of recurrent neural network (RNN) classifiers for precise discrimination of three-clinical states. Comparative analysis showed that the proposed Bi-directional long-short-term-memory (Bi-LSTM) model dominated the long-short-term-memory (LSTM) network by attaining an overall accuracy of 98.22% for the three-class classification task, whereas LSTM hardly achieved 94.22% accuracy on the test dataset.

Classification of Brain Tumor from Magnetic Resonance Imaging Using Vision Transformers Ensembling.

The automated classification of brain tumors plays an important role in supporting radiologists in decision making. Recently, vision transformer (ViT)-based deep neural network architectures have gained attention in the computer vision research domain owing to the tremendous success of transformer models in natural language processing. Hence, in this study, the ability of an ensemble of standard ViT models for the diagnosis of brain tumors from T1-weighted (T1w) magnetic resonance imaging (MRI) is investigated. Pretrained and finetuned ViT models (B/16, B/32, L/16, and L/32) on ImageNet were adopted for the classification task. A brain tumor dataset from figshare, consisting of 3064 T1w contrast-enhanced (CE) MRI slices with meningiomas, gliomas, and pituitary tumors, was used for the cross-validation and testing of the ensemble ViT model’s ability to perform a three-class classification task. The best individual model was L/32, with an overall test accuracy of 98.2% at 384 × 384 resolution. The ensemble of all four ViT models demonstrated an overall testing accuracy of 98.7% at the same resolution, outperforming individual model’s ability at both resolutions and their ensembling at 224 × 224 resolution. In conclusion, an ensemble of ViT models could be deployed for the computer-aided diagnosis of brain tumors based on T1w CE MRI, leading to radiologist relief.

Residual one-dimensional convolutional neural network for neuromuscular disorder classification from needle electromyography signals with explainability

Background and ObjectiveNeuromuscular disorders are diseases that damage our ability to control body movements. Needle electromyography (nEMG) is often used to diagnose neuromuscular disorders, which is an electrophysiological test measuring electric signals generated from a muscle using an invasive needle. Characteristics of nEMG signals are manually analyzed by an electromyographer to diagnose the types of neuromuscular disorders, and this process is highly dependent on the subjective experience of the electromyographer. Contemporary computer-aided methods utilized deep learning image classification models to classify nEMG signals which are not optimized for classifying signals. Additionally, model explainability was not addressed which is crucial in medical applications. This study aims to improve prediction accuracy, inference time, and explain model predictions in nEMG neuromuscular disorder classification. MethodsThis study introduces the nEMGNet, a one-dimensional convolutional neural network with residual connections designed to extract features from raw signals with higher accuracy and faster speed compared to image classification models from previous works. Next, the divide-and-vote (DiVote) algorithm was designed to integrate each subject's heterogeneous nEMG signal data structures and to utilize muscle subtype information for higher accuracy. Finally, feature visualization was used to identify the causality of nEMGNet diagnosis predictions, to ensure that nEMGNet made predictions on valid features, not artifacts. ResultsThe proposed method was tested using 376 nEMG signals measured from 57 subjects between June 2015 to July 2020 in Seoul National University Hospital. The results from the three-class classification task demonstrated that nEMGNet's prediction accuracy of nEMG signal segments was 62.35%, and the subject diagnosis prediction accuracy of nEMGNet and the DiVote algorithm was 83.69 %, over 5-fold cross-validation. nEMGNet outperformed all models from previous works on nEMG diagnosis classification, and heuristic analysis of feature visualization results indicate that nEMGNet learned relevant nEMG signal characteristics. ConclusionsThis study introduced nEMGNet and DiVote algorithm which demonstrated fast and accurate performance in predicting neuromuscular disorders based on nEMG signals. The proposed method may be applied in medicine to support real-time electrophysiologic diagnosis.

Distinguishing natural and computer generated images using Multi-Colorspace fused EfficientNet

The problem of distinguishing natural images from photo-realistic computer generated ones either addresses natural images versus computer graphics or natural images versus GAN images at a time. But in a real-world image forensic scenario, it is highly essential to consider all categories of image generation since in most cases image generation is unknown. We for the first time to our best knowledge, approach the problem of distinguishing natural images from photo-realistic computer generated images as a three-class classification task classifying natural, computer graphics and GAN images. For the task, we propose a Multi-Colorspace fused EfficientNet model by parallelly fusing three EfficientNet networks that follow transfer learning methodology where each of the three networks operates in a different colorspace, one in RGB, the other in LCH and the last in HSV that are chosen after analyzing the efficacy of various colorspace transformations in this image forensics problem. Our model outperforms the baselines in terms of accuracy, robustness towards post-processing and generalizability towards other datasets. We conduct psychophysics experiments to understand how accurately humans can distinguish natural, computer graphics and GAN images where we could observe that humans find difficulty in classifying these images, particularly the computer generated images, indicating the necessity of computational algorithms for the task. We also analyze the behavior of our model through visual explanations to understand salient regions that contribute to model’s decision making and compare with manual explanations provided by human participants in the form of region markings where we could observe similarities in both the explanations indicating powerful nature of our model to take the decisions meaningfully.

Diffusion MRI signal cumulants and hepatocyte microstructure at fixed diffusion time: Insights from simulations, 9.4T imaging, and histology.

PurposeRelationships between diffusion‐weighted MRI signals and hepatocyte microstructure were investigated to inform liver diffusion MRI modeling, focusing on the following question: Can cell size and diffusivity be estimated at fixed diffusion time, realistic SNR, and negligible contribution from extracellular/extravascular water and exchange? MethodsMonte Carlo simulations were performed within synthetic hepatocytes for varying cell size/diffusivity L/D0, and clinical protocols (single diffusion encoding; maximum b‐value: {1000, 1500, 2000} s/mm2; 5 unique gradient duration/separation pairs; SNR = {∞, 100, 80, 40, 20}), accounting for heterogeneity in (D0,L) and perfusion contamination. Diffusion (D) and kurtosis (K) coefficients were calculated, and relationships between (D0,L) and (D,K) were visualized. Functions mapping (D,K) to (D0,L) were computed to predict unseen (D0,L) values, tested for their ability to classify discrete cell‐size contrasts, and deployed on 9.4T ex vivo MRI‐histology data of fixed mouse liversResultsRelationships between (D,K) and (D0,L) are complex and depend on the diffusion encoding. Functions mapping D,K to (D0,L) captures salient characteristics of D0(D,K) and L(D,K) dependencies. Mappings are not always accurate, but they enable just under 70% accuracy in a three‐class cell‐size classification task (for SNR = 20, bmax = 1500 s/mm2, δ = 20 ms, and Δ = 75 ms). MRI detects cell‐size contrasts in the mouse livers that are confirmed by histology, but overestimates the largest cell sizes.ConclusionSalient information about liver cell size and diffusivity may be retrieved from minimal diffusion encodings at fixed diffusion time, in experimental conditions and pathological scenarios for which extracellular, extravascular water and exchange are negligible.

Application of Deep Convolutional Neural Networks for Discriminating Benign, Borderline, and Malignant Serous Ovarian Tumors From Ultrasound Images.

ObjectiveThis study aimed to evaluate the performance of the deep convolutional neural network (DCNN) to discriminate between benign, borderline, and malignant serous ovarian tumors (SOTs) on ultrasound(US) images.Material and MethodsThis retrospective study included 279 pathology-confirmed SOTs US images from 265 patients from March 2013 to December 2016. Two- and three-class classification task based on US images were proposed to classify benign, borderline, and malignant SOTs using a DCNN. The 2-class classification task was divided into two subtasks: benign vs. borderline & malignant (task A), borderline vs. malignant (task B). Five DCNN architectures, namely VGG16, GoogLeNet, ResNet34, MobileNet, and DenseNet, were trained and model performance before and after transfer learning was tested. Model performance was analyzed using accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve (AUC).ResultsThe best overall performance was for the ResNet34 model, which also achieved the better performance after transfer learning. When classifying benign and non-benign tumors, the AUC was 0.96, the sensitivity was 0.91, and the specificity was 0.91. When predicting malignancy and borderline tumors, the AUC was 0.91, the sensitivity was 0.98, and the specificity was 0.74. The model had an overall accuracy of 0.75 for in directly classifying the three categories of benign, malignant and borderline SOTs, and a sensitivity of 0.89 for malignant, which was better than the overall diagnostic accuracy of 0.67 and sensitivity of 0.75 for malignant of the senior ultrasonographer.ConclusionDCNN model analysis of US images can provide complementary clinical diagnostic information and is thus a promising technique for effective differentiation of benign, borderline, and malignant SOTs.

Automated classification of acute leukemia on a heterogeneous dataset using machine learning and deep learning techniques

Today, artificial intelligence and deep learning techniques constitute a prominent part in the area of medical sciences. These techniques help doctors detect diseases early and reduce their burden as well as chances of errors. However, experiments based on deep learning techniques require large and well-annotated dataset. This paper introduces a novel dataset of 500 peripheral blood smear images, containing normal, Acute Myeloid Leukemia and Acute Lymphoblastic Leukemia images. The dataset comprises almost 1700 cancerous blood cells. The size of the dataset is increased by adding images of a publicly available dataset and forming a heterogeneous dataset. The heterogeneous dataset is used for the automated binary classification task, which is one of the major tasks of the proposed work. The proposed work perform binary as well as three-class classification tasks involving state-of-the-art techniques based on machine learning and deep learning. For binary classification, the proposed work achieved an accuracy of 97% when fully connected layers along with the last three convolutional layers of VGG16 are fine tuned and 98% for DenseNet121 along with support vector machine. For three-class classification task, an accuracy of 95% is obtained for ResNet50 along with support vector machine. The preparation of the novel dataset is done under the opinion of various expertise that will help the scientific community for medical research supported by machine learning models.

Machine Learning Based Metagenomic Prediction of Inflammatory Bowel Disease.

In this study, we investigate faecal microbiota composition, in an attempt to evaluate performance of classification algorithms in identifying Inflammatory Bowel Disease (IBD) and its two types: Crohn's disease (CD) and ulcerative colitis (UC). From many investigated algorithms, a random forest (RF) classifier was selected for detailed evaluation in three-class (CD versus UC versus nonIBD) classification task and two binary (nonIBD versus IBD and CD versus UC) classification tasks. We dealt with class imbalance, performed extensive parameter search, dimensionality reduction and two-level classification. In three-class classification, our best model reaches F1 score of 91% in average, which confirms the strong connection of IBD and gastrointestinal microbiome. Among most important features in three-class classification are species Staphylococcus hominis, Porphyromonas endodontalis, Slackia piriformis and genus Bacteroidetes.

COVID-Nets: deep CNN architectures for detecting COVID-19 using chest CT scans.

In this paper we propose two novel deep convolutional network architectures, CovidResNet and CovidDenseNet, to diagnose COVID-19 based on CT images. The models enable transfer learning between different architectures, which might significantly boost the diagnostic performance. Whereas novel architectures usually suffer from the lack of pretrained weights, our proposed models can be partly initialized with larger baseline models like ResNet50 and DenseNet121, which is attractive because of the abundance of public repositories. The architectures are utilized in a first experimental study on the SARS-CoV-2 CT-scan dataset, which contains 4173 CT images for 210 subjects structured in a subject-wise manner into three different classes. The models differentiate between COVID-19, non-COVID-19 viral pneumonia, and healthy samples. We also investigate their performance under three binary classification scenarios where we distinguish COVID-19 from healthy, COVID-19 from non-COVID-19 viral pneumonia, and non-COVID-19 from healthy, respectively. Our proposed models achieve up to 93.87% accuracy, 99.13% precision, 92.49% sensitivity, 97.73% specificity, 95.70% F1-score, and 96.80% AUC score for binary classification, and up to 83.89% accuracy, 80.36% precision, 82.04% sensitivity, 92.07% specificity, 81.05% F1-score, and 94.20% AUC score for the three-class classification tasks. We also validated our models on the COVID19-CT dataset to differentiate COVID-19 and other non-COVID-19 viral infections, and our CovidDenseNet model achieved the best performance with 81.77% accuracy, 79.05% precision, 84.69% sensitivity, 79.05% specificity, 81.77% F1-score, and 87.50% AUC score. The experimental results reveal the effectiveness of the proposed networks in automated COVID-19 detection where they outperform standard models on the considered datasets while being more efficient.

CovidXrayNet: Optimizing data augmentation and CNN hyperparameters for improved COVID-19 detection from CXR

To mitigate the spread of the current coronavirus disease 2019 (COVID-19) pandemic, it is crucial to have an effective screening of infected patients to be isolated and treated. Chest X-Ray (CXR) radiological imaging coupled with Artificial Intelligence (AI) applications, in particular Convolutional Neural Network (CNN), can speed the COVID-19 diagnostic process. In this paper, we optimize the data augmentation and the CNN hyperparameters for detecting COVID-19 from CXRs in terms of validation accuracy. This optimization increases the accuracy of the popular CNN architectures such as the Visual Geometry Group network (VGG-19) and the Residual Neural Network (ResNet-50), by 11.93% and 4.97%, respectively. We then proposed CovidXrayNet model that is based on EfficientNet-B0 and our optimization results. We evaluated CovidXrayNet on two datasets, including our generated balanced COVIDcxr dataset (960 CXRs) and the benchmark COVIDx dataset (15,496 CXRs). With only 30 epochs of training, CovidXrayNet achieves state-of-the-art accuracy of 95.82% on the COVIDx dataset in the three-class classification task (COVID-19, normal or pneumonia). The CovidXRayNet model, the COVIDcxr dataset, and several optimization experiments are publicly available at https://github.com/MaramMonshi/CovidXrayNet.

Matrix Relevance Learning From Spectral Data for Diagnosing Cassava Diseases

We discuss the use of matrix relevance learning, a popular extension to prototype learning algorithms, applied to a three-class classification task of diagnosing cassava diseases from spectral data. Previously this diagnosis has been done using plant image data taken with a smartphone. However for this method disease symptoms need to be visible. Unfortunately for some cassava diseases, once symptoms have manifested on the aerial part of the plant, the root which is the edible part of the plant has been totally destroyed. This research is premised on the hypothesis that diseased crops without visible symptoms can be detected using spectral information, allowing for early interventions. In this paper, we analyze visible and near-infrared spectra captured from leaves infected with two common cassava diseases (cassava brown streak disease and cassava mosaic virus disease) found in Sub-Saharan Africa. We also take spectra from leaves of healthy plants. The spectral data come with thousands of dimensions, therefore different wavelengths are analyzed in order to identify the most relevant spectral bands for diagnosing these disease. To cope with the nominally high number of input dimensions of data, functional decomposition of the spectra is applied. The classification task is addressed using Generalized Matrix Relevance Learning Vector Quantization and compared with the standard classification techniques performed in the space of expansion coefficients.