Comparative Analysis of Conditional Deep Convolutional and Wasserstein GAN Architectures for Brain Tumor MRI Data Augmentation

DOI

10.2991/978-94-6463-948-3_15How to use a DOI?

Keywords

Generative Adversarial Networks; cGAN; DCGAN; WGAN; Brain MRI; Medical Image Augmentation

Abstract

Brain MRI datasets with detailed annotations are hard to collect, which limits the effectiveness of diagnostic models. Brain tumor images are especially challenging, and available datasets are often small, which restricts training. We test whether synthetic images can expand these datasets in a useful way. We compare three GAN models, conditional GAN, DCGAN, and WGAN, and train them on a balanced sample of 400 MRI scans with 200 tumor and 200 healthy images. All models follow the same training setup. We measure image quality and diversity using the Inception Score and the Fréchet Inception Distance. The conditional GAN produces the strongest results. It reaches an Inception Score of 2.035 and a FID of 113.81. DCGAN generates average quality, and WGAN offers stable training with lower detail. These results support the use of conditional GANs for class specific medical image augmentation.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yogita D. Patil
AU  - Karthik Kurup
AU  - Sadiq Shaikh
AU  - Nitiraj V. Kulkarni
PY  - 2026
DA  - 2026/01/06
TI  - Comparative Analysis of Conditional Deep Convolutional and Wasserstein GAN Architectures for Brain Tumor MRI Data Augmentation
BT  - Proceedings of the International Conference on Sustainable Innovation with Artificial Intelligence and Machine Learning 2025 (ICSIAIML 2025)
PB  - Atlantis Press
SP  - 223
EP  - 233
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6463-948-3_15
DO  - 10.2991/978-94-6463-948-3_15
ID  - Patil2026
ER  -