Attention mechanism and mixup data augmentation for classification of COVID-19 Computed Tomography images

A Hybrid Deep Transfer Learning Framework for Delamination Identification in Composite Laminates

Structural health monitoring (SHM) has proven to be an effective technique to maintain the safety and reliability of laminated composites. Recently, both deep learning and machine learning methodologies have gained popularity in sensor-based SHM. However, machine learning approaches often require tedious manual feature extraction, while deep learning models require large training datasets, which may not be feasible. To overcome these limitations, this study presents a hybrid deep transfer learning (HTL) framework to identify delamination in composite laminates. The proposed framework enhances SHM performance by utilizing pre-trained EfficientNet and ResNet models to allow for deep feature extraction with limited data. EfficientNet contributes to this by efficiently scaling the model to capture multi-scale spatial features, while ResNet contributes by extracting hierarchical representations through its residual connections. Vibration signals from piezoelectric (PZT) sensors attached to the composite laminates, consisting of three health states, are used to validate the approach. Compared to the existing transfer learning approaches, the suggested method achieved better performance, hence improving both the accuracy and robustness of delamination detection in composite structures.

Class-Wise Combination of Mixture-Based Data Augmentation for Class Imbalance Learning of Focal Liver Lesions in Abdominal CT Images

In this paper, we propose a method to address the class imbalance learning in the classification of focal liver lesions (FLLs) from abdominal CT images. Class imbalance is a significant challenge in medical image analysis, making it difficult for machine learning models to learn to classify them accurately. To overcome this, we propose a class-wise combination of mixture-based data augmentation (CCDA) method that uses two mixture-based data augmentation techniques, MixUp and AugMix. These are applied at different ratios for each class to adaptively learn features for each class. This method is tailored to handle the unique characteristics of each FLL type by adjusting the augmentation mix for major classes (e.g., cysts and metastases) and minor classes (e.g., hemangiomas). In experiments, our method was validated on the dataset consisted of portal phase CT images from 1290 colorectal cancer patients. The results showed that applying MixUp and AugMix in a class-wise manner could significantly improve the classification performance of minor classes while maintaining or slightly improving the performance of major classes. Quantitative results showed higher F1 scores for minor classes and a more balanced accuracy across classes when CCDA was used compared to other methods.

Developing precision agriculture using data augmentation framework for automatic identification of castor insect pests

Castor (Ricinus communis L.) is an important nonedible industrial crop that produces oil, which is used in the production of medicines, lubricants, and other products. However, the quality and quantity of castor oil are critical factors that can be degraded by various insect pest attacks. The traditional method of identifying the correct category of pests required a significant amount of time and expertise. To solve this issue, automatic insect pest detection methods combined with precision agriculture can help farmers in providing adequate support for sustainable agriculture development. For accurate predictions, the recognition system requires a sufficient amount of data from a real-world situation, which is not always available. In this regard, data augmentation is a popular technique used for data enrichment. The research conducted in this investigation established an insect pest dataset of common castor pests. This paper proposes a hybrid manipulation-based approach for data augmentation to solve the issue of the lack of a suitable dataset for effective vision-based model training. The deep convolutional neural networks VGG16, VGG19, and ResNet50 are then adopted to analyze the effects of the proposed augmentation method. The prediction results show that the proposed method addresses the challenges associated with adequate dataset size and significantly improves overall performance when compared to previous methods.

EVAE-Net: An Ensemble Variational Autoencoder Deep Learning Network for COVID-19 Classification Based on Chest X-ray Images

The COVID-19 pandemic has had a significant impact on many lives and the economies of many countries since late December 2019. Early detection with high accuracy is essential to help break the chain of transmission. Several radiological methodologies, such as CT scan and chest X-ray, have been employed in diagnosing and monitoring COVID-19 disease. Still, these methodologies are time-consuming and require trial and error. Machine learning techniques are currently being applied by several studies to deal with COVID-19. This study exploits the latent embeddings of variational autoencoders combined with ensemble techniques to propose three effective EVAE-Net models to detect COVID-19 disease. Two encoders are trained on chest X-ray images to generate two feature maps. The feature maps are concatenated and passed to either a combined or individual reparameterization phase to generate latent embeddings by sampling from a distribution. The latent embeddings are concatenated and passed to a classification head for classification. The COVID-19 Radiography Dataset from Kaggle is the source of chest X-ray images. The performances of the three models are evaluated. The proposed model shows satisfactory performance, with the best model achieving 99.19% and 98.66% accuracy on four classes and three classes, respectively.