Super-Resolution with RCAN for Improved Malaria Cell Classification: A Performance Evaluation

This study investigates the use of Residual Channel Attention Networks (RCAN) for $2 x$ super-resolution in malaria cell imaging to enhance the diagnostic capabilities of image classification models. While foundational for malaria diagnosis, traditional microscopy techniques often yield variable results contingent upon the operator’s proficiency. RCAN was employed to upscale the resolution of malaria cell images, allowing for the extraction of more detailed and detectable features. Super- resolution effectiveness was quantitatively assessed using the Peak Signal Noise Ratio (PSNR) and Structural Similarity Index Mea- sure (SSIM). Post-super-resolution, a comprehensive horizontal comparison of multiple classification models was conducted using both the original and enhanced datasets to demonstrate the superiority of super-resolved images in improving classification accuracy. The model that exhibited the best performance was further analyzed to provide interpretability insights through confusion matrix, precision-recall curves, ROC analysis, F1- Score, and visualization of semantic features. This analysis not only underscored the enhanced accuracy