Comparative Analysis of VGG16 and InceptionV3 for Multi-Label Chest X-Ray Classification

This paper employs the NIH ChestX-ray14 dataset to compare the VGG16 and InceptionV3 architectures for multi-label chest X-ray classification. The impact of CLAHE-based contrast enhancement on convolutional neural network performance is investigated. Transfer learning was applied using ImageNet-pretrained weights. ROC curves and AUC metrics were used to assess classification performance, highlighting the influence of preprocessing and architectural design on medical image analysis. The results show that InceptionV3 outperforms VGG16, while CLAHE-based preprocessing improves the classification performance of both architectures.