Background: The development of advanced computational models for medical imaging is crucial for improving diagnostic accuracy in healthcare. This paper introduces a novel approach for virtual contrast enhancement (VCE) in magnetic resonance imaging (MRI), particularly focusing on nasopharyngeal cancer (NPC).Methods: The proposed model, Pixelwise Gradient Model with GAN for Virtual Contrast Enhancement (PGMGVCE), makes use of pixelwise gradient methods with Generative Adversarial Networks (GANs) to enhance T1-weighted (T1-w) and T2-weighted (T2-w) MRI images. This approach combines the benefits of both modalities to simulate the effects of gadolinium-based contrast agents, thereby reducing associated risks. Various modifications of PGMGVCE, including changing hyperparameters, using normalization methods (z-score, Sigmoid and Tanh) and training the model with T1-w or T2-w images only, were tested to optimize the model’s performance.Results: PGMGVCE demonstrated a similar accuracy to the existing model in terms of mean absolute error (MAE) (8.56±0.45 for Li’s model; 8.72±0.48 for PGMGVCE), mean square error (MSE) (12.43±0.67 for Li’s model; 12.81±0.73 for PGMGVCE) and structural similarity index (SSIM) (0.71±0.08 for Li’s model; 0.73±0.12 for PGMGVCE). However, it showed improvements in texture representation, as indicated by total mean square variation per mean intensity (TMSVPMI) (0.124±0.022 for ground truth; 0.079±0.024 for Li’s model; 0.120±0.027 for PGMGVCE), total absolute variation per mean intensity (TAVPMI) (0.159±0.031 for ground truth; 0.100±0.032 for Li’s model; 0.153±0.029 for PGMGVCE), Tenengrad function per mean intensity (TFPMI) (1.222±0.241 for ground truth; 0.981±0.213 for Li’s model; 1.194±0.223 for PGMGVCE) and variance function per mean intensity (VFPMI) (0.0811±0.005 for ground truth; 0.0667±0.006 for Li’s model; 0.0761±0.006 for PGMGVCE).Conclusions: PGMGVCE presents an innovative and safe approach to VCE in MRI, demonstrating the power of deep learning in enhancing medical imaging. This model paves the way for more accurate and risk-free diagnostic tools in medical imaging.
背景：开发先进的医学影像计算模型对提升医疗诊断准确性至关重要。本文提出一种磁共振成像（MRI）虚拟对比增强（VCE）的新方法，特别聚焦于鼻咽癌（NPC）的影像分析。 方法：所提出的模型——基于生成对抗网络的像素梯度虚拟对比增强模型（PGMGVCE），利用像素梯度方法与生成对抗网络（GAN）增强T1加权（T1-w）和T2加权（T2-w）MRI图像。该方法综合两种模态的优势，模拟钆基对比剂的效果，从而降低相关风险。为优化模型性能，测试了PGMGVCE的多种改进方案，包括调整超参数、采用归一化方法（z-score、Sigmoid和Tanh）以及仅使用T1-w或T2-w图像进行训练。 结果：PGMGVCE在平均绝对误差（MAE）（Li模型：8.56±0.45；PGMGVCE：8.72±0.48）、均方误差（MSE）（Li模型：12.43±0.67；PGMGVCE：12.81±0.73）和结构相似性指数（SSIM）（Li模型：0.71±0.08；PGMGVCE：0.73±0.12）方面与现有模型精度相当。但在纹理表征方面表现更优，具体指标包括：单位平均强度的总均方变异（TMSVPMI）（真实值：0.124±0.022；Li模型：0.079±0.024；PGMGVCE：0.120±0.027）、单位平均强度的总绝对变异（TAVPMI）（真实值：0.159±0.031；Li模型：0.100±0.032；PGMGVCE：0.153±0.029）、单位平均强度的Tenengrad函数（TFPMI）（真实值：1.222±0.241；Li模型：0.981±0.213；PGMGVCE：1.194±0.223）以及单位平均强度的方差函数（VFPMI）（真实值：0.0811±0.005；Li模型：0.0667±0.006；PGMGVCE：0.0761±0.006）。 结论：PGMGVCE为MRI虚拟对比增强提供了一种创新且安全的方法，展现了深度学习在增强医学影像方面的潜力。该模型为开发更精准、无风险的医学影像诊断工具开辟了新途径。
Pixelwise Gradient Model with GAN for Virtual Contrast Enhancement in MRI Imaging
肿瘤(癌症)患者之家