MAN-GAN: a mask-adaptive normalization based generative adversarial networks for liver multi-phase CT image...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Liver multiphase enhanced computed tomography (MPECT) is crucial for diagnosis but faces limitations.
• Generating MPECT images typically requires contrast agents, posing risks and limitations.

Purpose of the Study:

• To develop a deep learning network for automatic MPECT image generation from non-contrast CT scans.
• To evaluate the performance and clinical utility of the proposed deep learning model.

Main Methods:

• A Mask-Adaptive Normalization-based Generative Adversarial Network with Cycle-Consistency Loss (MAN-GAN) was developed.
• The model was trained and validated on multiple datasets, including internal and external validation sets.
• Comparative analysis with state-of-the-art methods and radiologist evaluations were conducted.

Main Results:

• MAN-GAN outperformed baseline and other state-of-the-art methods in generating MPECT images across all three phases.
• Generated images demonstrated above-average quality and satisfactory similarity to real MPECT images.
• Performance was validated across internal and external datasets.

Conclusions:

• MAN-GAN successfully translates non-contrast CT images to MPECT images, demonstrating feasibility and state-of-the-art performance.
• The model offers a potential solution for contrast-related limitations in liver CT imaging.
• This approach has significant potential to aid clinical scenarios involving liver CT analysis.