Explainable Deep Learning in Medical Imaging: Brain Tumor and Pneumonia Detection

TL;DR

The study tackles the interpretability gap in deep learning for medical imaging by evaluating ResNet50 and DenseNet121 on brain tumor detection in MRI and pneumonia detection in chest X-rays using Grad-CAM explanations. Through ImageNet-pretrained transfer learning, data augmentation, and a consistent training/evaluation protocol on public Kaggle datasets, DenseNet121 outperforms ResNet50 in both accuracy and AUC, while also delivering more clinically meaningful Grad-CAM visualizations. The results demonstrate that high-performing architectures paired with explainability tools can produce reliable, transparent diagnostic aids. This work highlights the practical potential of explainable deep learning in real-world clinical workflows and motivates further expansion to additional diseases and advanced XAI techniques.

Abstract

Deep Learning (DL) holds enormous potential for improving medical imaging diagnostics, yet the lack of interpretability in most models hampers clinical trust and adoption. This paper presents an explainable deep learning framework for detecting brain tumors in MRI scans and pneumonia in chest X-ray images using two leading Convolutional Neural Networks, ResNet50 and DenseNet121. These models were trained on publicly available Kaggle datasets comprising 7,023 brain MRI images and 5,863 chest X-ray images, achieving high classification performance. DenseNet121 consistently outperformed ResNet50 with 94.3 percent vs. 92.5 percent accuracy for brain tumors and 89.1 percent vs. 84.4 percent accuracy for pneumonia. For better explainability, Gradient-weighted Class Activation Mapping (Grad-CAM) was integrated to create heatmap visualizations superimposed on the test images, indicating the most influential image regions in the decision-making process. Interestingly, while both models produced accurate results, Grad-CAM showed that DenseNet121 consistently focused on core pathological regions, whereas ResNet50 sometimes scattered attention to peripheral or non-pathological areas. Combining deep learning and explainable AI offers a promising path toward reliable, interpretable, and clinically useful diagnostic tools.

Figures (7)

• Figure 1: Sample images from the Brain MRI dataset showing tumor, and no tumor cases.
• Figure 2: Sample images from the Chest X-ray dataset showing Pneumonia and normal cases.
• Figure 3: High-level architecture of the ResNet50 deep neural network used for both brain MRI and Pneumonia classification tasks.
• Figure 4: High-level architecture of DenseNet121 model illustrating dense connectivity and feature reuse across layers.
• Figure 5: ROC curves illustrating the performance of DenseNet models on brain tumor (left) and pneumonia (right) classification tasks.