Adaptive Input-image Normalization for Solving the Mode Collapse Problem in GAN-based X-ray Images

TL;DR

This work tackles data imbalance and GAN mode collapse in chest X-ray datasets by introducing adaptive input-image normalization (AIIN) integrated with DCGAN and ACGAN. AIIN preprocesses X-ray images via contrast-based histogram normalization to emphasize diverse anatomical features, improving intra-class and inter-class diversity of synthetic images as measured by MS-SSIM, FID, and IS, respectively. The augmented data are used to train a Vision Transformer classifier, yielding mixed but often improved classification metrics, particularly for Pneumonia, while highlighting modality-dependent limitations and the need for domain-specific evaluation. Overall, AIIN provides a practical preprocessing-based boost to GAN-based data augmentation in biomedical imaging, with potential extensions to more advanced GANs and additional imaging modalities.

Abstract

Biomedical image datasets can be imbalanced due to the rarity of targeted diseases. Generative Adversarial Networks play a key role in addressing this imbalance by enabling the generation of synthetic images to augment datasets. It is important to generate synthetic images that incorporate a diverse range of features to accurately represent the distribution of features present in the training imagery. Furthermore, the absence of diverse features in synthetic images can degrade the performance of machine learning classifiers. The mode collapse problem impacts Generative Adversarial Networks' capacity to generate diversified images. Mode collapse comes in two varieties: intra-class and inter-class. In this paper, both varieties of the mode collapse problem are investigated, and their subsequent impact on the diversity of synthetic X-ray images is evaluated. This work contributes an empirical demonstration of the benefits of integrating the adaptive input-image normalization with the Deep Convolutional GAN and Auxiliary Classifier GAN to alleviate the mode collapse problems. Synthetically generated images are utilized for data augmentation and training a Vision Transformer model. The classification performance of the model is evaluated using accuracy, recall, and precision scores. Results demonstrate that the DCGAN and the ACGAN with adaptive input-image normalization outperform the DCGAN and ACGAN with un-normalized X-ray images as evidenced by the superior diversity scores and classification scores.

Figures (10)

• Figure 1: Proposed methodology for data augmentation of chest X-ray datasets with GAN-based generated images.
• Figure 2: Block diagram of DCGAN architecture saad2022addressing.
• Figure 3: Block diagram of ACGAN architecture.
• Figure 4: COVID-19 and Pneumonia real and synthetic samples from the un-normalized and AIIN normalized X-ray images.
• Figure 5: The MS-SSIM and FID scores of DCGAN for the un-normalized and AIIN normalized COVID-19 X-ray images.