Acute Lymphoblastic Leukemia (ALL) is a blood cancer requiring swift, accurate diagnosis for effective treatment. Traditional diagnostic methods rely on the manual examination of microscopic images by experts, which is limited by both availability and speed. Artificial Intelligence (AI) offers a promising alternative for the automated analysis of blood cell images. However, AI model development is hindered by the need for large, annotated datasets, which are often scarce and exhibit a bias (more images of healthy cells than ALL samples). This research proposes the use of data augmentation techniques, specifically Generative Adversarial Networks (GANs), to mitigate these issues. GANs are capable of producing realistic synthetic images of ALL cells, which can enhance the dataset and correct the imbalance. This improvement can lead to the creation of more accurate AI models for ALL diagnosis, potentially improving patient outcomes. The study evaluates three GAN architectures namely SimpleGAN, Wasserstein GAN (WGAN) and Deep Convolutional GAN (DCGAN) for their ability to generate synthetic images of ALL and normal blood cells. The performance is measured by analyzing loss metrics and accuracy rates during training to identify the optimal GAN architecture for this application. Results indicate that DCGAN significantly outperforms both SimpleGAN and WGAN. A higher Peak Signal-to-Noise Ratio (PSNR) value implies that images generated by DCGAN have less noise, and a higher Structural Similarity Index Measure (SSIM) value indicates a better preservation of structural details in the images.