Zafar Junaid, Koc Vincent, Zafar Haroon
Faculty of Engineering, Government College University, Lahore 54000, Pakistan.
Independent Researcher, Kuala Lumpur 55100, Malaysia.
J Imaging. 2025 Mar 28;11(4):101. doi: 10.3390/jimaging11040101.
Generative adversarial networks (GANs) prioritize pixel-level attributes over capturing the entire image distribution, which is critical in image synthesis. To address this challenge, we propose a dual-stream contrastive latent projection generative adversarial network (DSCLPGAN) for the robust augmentation of MRI images. The dual-stream generator in our architecture incorporates two specialized processing pathways: one is dedicated to local feature variation modeling, while the other captures global structural transformations, ensuring a more comprehensive synthesis of medical images. We used a transformer-based encoder-decoder framework for contextual coherence and the contrastive learning projection (CLP) module integrates contrastive loss into the latent space for generating diverse image samples. The generated images undergo adversarial refinement using an ensemble of specialized discriminators, where discriminator 1 (D1) ensures classification consistency with real MRI images, discriminator 2 (D2) produces a probability map of localized variations, and discriminator 3 (D3) preserves structural consistency. For validation, we utilized a publicly available MRI dataset which contains 3064 T1-weighted contrast-enhanced images with three types of brain tumors: meningioma (708 slices), glioma (1426 slices), and pituitary tumor (930 slices). The experimental results demonstrate state-of-the-art performance, achieving an SSIM of 0.99, classification accuracy of 99.4% for an augmentation diversity level of 5, and a PSNR of 34.6 dB. Our approach has the potential of generating high-fidelity augmentations for reliable AI-driven clinical decision support systems.
生成对抗网络（GAN）在图像合成中，将像素级属性置于捕捉整个图像分布之上，而捕捉整个图像分布对图像合成至关重要。为应对这一挑战，我们提出了一种双流对比潜在投影生成对抗网络（DSCLPGAN），用于对MRI图像进行稳健增强。我们架构中的双流生成器包含两条专门的处理路径：一条致力于局部特征变化建模，另一条捕捉全局结构变换，以确保对医学图像进行更全面的合成。我们使用基于Transformer的编码器 - 解码器框架来实现上下文连贯性，并且对比学习投影（CLP）模块将对比损失整合到潜在空间中，以生成多样化的图像样本。生成的图像使用一组专门的鉴别器进行对抗性细化，其中鉴别器1（D1）确保与真实MRI图像的分类一致性，鉴别器2（D2）生成局部变化的概率图，鉴别器3（D3）保持结构一致性。为了进行验证，我们使用了一个公开可用的MRI数据集，该数据集包含3064张T1加权对比增强图像，其中有三种类型的脑肿瘤：脑膜瘤（708个切片）、胶质瘤（1426个切片）和垂体瘤（930个切片）。实验结果表明，该方法具有先进的性能，对于增强多样性水平为5的情况，达到了0.99的结构相似性指数（SSIM）、99.4%的分类准确率和34.6 dB的峰值信噪比（PSNR）。我们的方法有潜力为可靠的人工智能驱动的临床决策支持系统生成高保真增强图像。
