Abstract
Abstract
Medical images are valuable resources for clinical diagnosis, each containing rich information about diseases. To fully utilize this information, it is crucial to achieve effective conversion between different types of medical images. This not only reduces the need for patients to undergo multiple imaging examinations but also helps to conserve valuable medical resources and reduce examination restrictions. Therefore, research on medical image transformation techniques is of significant importance, but it is also a highly challenging task. This study proposes an innovative medical image generation network based on the U-Net architecture, specifically targeting brain CT and MRI images. Through carefully designed preprocessing steps and pretraining strategies, the network in this paper can effectively convert brain CT images into MRI images, demonstrating significant performance on the SynthRAD2023 Grand Challenge dataset. Additionally, in clinical applications, MRI is favored for its high-resolution soft tissue imaging capabilities, while CT scans are known for providing high-resolution images quickly. However, the MRI scanning process is longer and may not be suitable for patients who cannot maintain specific positions for a long time. Therefore, the conversion technique between CT and MRI images plays a crucial role in improving the efficiency of medical image acquisition. The results of this study will enhance the work of converting CT to MRI images in clinical practice.