Aug 2026· Algorithms· Vol 19, pp. 715· 0 citations· 49 references
TL;DR
This paper proposes ResKAN18, a hybrid structure that embeds the learnable spline function of the Kolmogorov–Arnold network (KAN) into the ResNet18 classification head for intelligent diagnosis of COVID-19 in chest X-ray images.
Abstract
In response to the challenge of balancing accuracy and generalization in medical image classification using convolutional neural networks (CNNs), this paper proposes ResKAN18, a hybrid structure that embeds the learnable spline function of the Kolmogorov–Arnold network (KAN) into the ResNet18 classification head for intelligent diagnosis of COVID-19 in chest X-ray images. ResKAN18 includes three variants: ResKAN18—Large (four layers of KAN, hidden-layer dimensions [256, 128, 64]), ResKAN18—Standard (four layers of KAN, hidden-layer dimensions [128, 64, 32]), and ResKAN18—Simple (three layers of KAN, hidden-layer dimensions [64, 32]), which can achieve a flexible balance between accuracy and efficiency with different depths of KAN. A systematic comparison has been conducted between four classic CNN baselines including ResNet18, VGG16, DenseNet121, ShuffleNetV2, and three ResKAN18 variants on a benchmark dataset containing 3880 chest X-rays (COVID-19, normal, viral pneumonia). The results have shown that ResKAN18—Large can achieve an accuracy of 98.80% on the independent test set, which is 1.21% higher than ResNet18 and 0.69% higher than DenseNet121—its parameter count is 13.97M, inference delay is 8.25 ms/image, and training–validation accuracy difference is only 1.50%. The accuracy and performance stability of the dataset under random partitioning conditions are superior to the other two variants and all classic CNN baselines. All ResKAN18 variants have achieved zero missed diagnoses for COVID-19, while ResNet18 has shown missed diagnoses (0.9944). Taking into account the trade-off between accuracy, generalization, and inference efficiency, ResKAN1—Large is recommended as the default configuration, while for edge deployment scenarios with severely limited resources, ResKAN18—Simple can provide a cost-effective alternative with an extremely low latency of 2.50 ms/image and only 3.1% parameter increment compared to ResNet18.
The results showed that residual learning on ResNet50 effectively addressed the vanishing gradient problem in deep networks and achieved clinically relevant classification accuracy, supporting its potential integration into computer-aided diagnosis (CAD) systems.
Gallen Cakra, Adhi Wibowo, Dita Madonna Simanjuntak et al.· International Journal of Hea...· 0 citations
A hybrid ensemble learning approach to classify chest X-ray images into four classes—Normal, COVID-19, Pneumonia, and Tuberculosis exhibited high sensitivity in detecting Tuberculosis with considerable stability in classifying Normal, COVID-19, and Pneumonia.
Abdul Rehman Khan Tareen, Muhammad Laiq Ur Rahman Shahid, Muhammad Hamza Zafar et al.· Allied Medical Research Jour...· 0 citations
Because the application of data augmentation and regularization techniques, the training and validation loss curves of MobileNetV2 con-verged effectively, which in turn suppressed the issue of overfitting, confirming the superiority of MobileNetV2 in end-to-end feature extraction for COVID-19 imaging.
Haodong Yao· Proceedings of the 3rd Inter...· 0 citations
These days, digital pathology plays a crucial role in tumor detection and prognosis, particularly in objectively evaluating histological images for colorectal cancer (CRC). Deep learning models have demonstrated significant success in image classification, prompting researchers to adopt these methods for medical image analysis. This research explores deep learning methods, specifically using ResNet architectures (ResNet-18, ResNet-50, ResNet-101), combined with various optimization methods, including Adam, Stochastic Gradient Descent with Momentum (SGDM), and Root Mean Square Propagation (RMSProp), for classifying colorectal cancer types from histological images. A comprehensive evaluation was conducted to identify the most effective neural network architecture and optimization approach. Among the tested combinations, the ResNet-50 model with the Adam optimization method achieved the highest accuracy of 99.86% on the NCT-CRC-HE-7K dataset and an accuracy of 98.36% on the NCT-CRC-HE-100K dataset, outperforming other tested combinations and existing approaches in the literature, effectively differentiating between benign and aggressive colorectal cancer. Lastly, the suggested model was compared to the current models, which were assessed to ascertain the most effective neural network models and the best training approach for our case study on colon tumor segmentation.
H. K. Omer, Salwa M. Hasan, Lozan M. Abdullrahman et al.· passer of basic and applied...· 0 citations
Attention mechanisms are widely adopted in convolutional neural networks to improve feature representation, yet their consistent effectiveness in medical image classification remains uncertain. This study systematically evaluates channel and spatial attention mechanisms integrated into a ResNet50 architecture for chest X-ray classification tasks. Three attention configurations are examined, namely Squeeze-and-Excitation (SE), Convolutional Block Attention Module (CBAM), and a hybrid combination of both, using three publicly available chest X-ray datasets: Pneumonia Chest X-ray, COVID-QU-Ex, and Tuberculosis Chest X-ray. All model variants are trained under identical preprocessing and optimization settings to ensure a controlled and fair comparison. Each experiment is repeated across multiple random seeds to reduce the influence of stochastic variability, and final performance is reported as the mean across runs. Model effectiveness is assessed using accuracy, precision, recall, and F1 score, followed by paired t-tests to determine the statistical significance of observed differences. Quantitative results indicate that attention-augmented models produce marginal performance changes relative to the ResNet50 baseline across all three datasets. Statistical testing confirms that none of the observed improvements reach significance, with all p-values exceeding 0.05. These findings suggest that attention mechanisms do not function as universally effective performance boosters in chest X-ray classification. Instead, their role appears to be supportive in nature, contributing to feature representation stability rather than driving substantial predictive gains. Dataset characteristics and attention placement are identified as key factors influencing the degree of benefit observed.
Rapid and reliable diagnosis of COVID-19 is of fundamental importance for effective pandemic control, and chest X-ray imaging combined with deep learning provides a very useful tool for large-scale screening. Therefore, this paper presents a tri-class classification framework for distinguishing COVID-19, viral pneumonia, and normal cases from chest X-rays. The method is built on a pre-trained ResNet50 and incorporates CBAM attention modules at two well-justified stages: spatial attention in Layer3 for abnormal region localization and channel attention in Layer4 for semantic feature selection. More importantly, it addresses the class imbalance common in medical data by using a weighted cross-entropy loss function. Experiments on a public COVID-19 chest X-ray dataset demonstrate that the full model attains 90.74% precision and 74.65% F1-score, both superior to baseline methods. Ablation studies rigorously validate each component, and Precision-Recall curve analysis gives an AUC-PR of 0.942. From the results for the COVID-19 class it is clearly seen that the attention map visualization shows where the model is looking at clinically relevant lung regions, and since the inference time is 2.64 ms per image (379.1 FPS), the proposed framework thus achieves a good balance between accuracy and speed for clinical use.
Weizhen Yu· International Conference on...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
