An Explainable Contrastive-based Dilated Convolutional Network with Transformer for Pediatric Pneumonia Detection

Abstract

Pediatric pneumonia remains a significant global threat, surpassing all other communicable diseases in child mortality and necessitating rapid diagnosis. UNICEF identifies it as a primary cause of death in children under five. However, traditional image processing methods are time-consuming and struggle to capture distinct features due to low-intensity radiation in chest radiographs. Furthermore, skewed training data complicates the extraction of reliable and relevant features. To this end, we propose a novel EXplainable Contrastive-based Dilated Convolutional Network with Transformer (XCCNet) for pediatric pneumonia detection. XCCNet harnesses the spatial power of dilated convolutions and the global insights of contrastive-based transformers for effective feature refinement. A robust chest X-ray processing module tackles low-intensity radiographs, while adversarial-based data augmentation mitigates data distribution challenges. Furthermore, we actively integrate an explainability approach through feature visualization, directly aligning it with the attention region that pinpoints the presence of pneumonia or normality in radiographs. We comprehensively evaluate the efficacy of XCCNet on four publicly available datasets: Kermany, VinDR-PCXR, NIH-Pediatric, and Trivedi, achieved accuracies of 99.76%, 91.56%, 92.87%, and 97.19%, respectively. Quantitative and qualitative assessments demonstrate the effectiveness of XCCNet capabilities. XCCNet’s superior performance enhances the early diagnosis of pneumonia, potentially saving many children’s lives.