Zhao Jufeng | 电子与技术硕士

Adaptive Phase Image Denoising to Improve MRgFUS Thermometry with a Thermal-Response Gaussian Model

Wed, 01 Jan 2025 00:00:00 +0000

Magnetic Resonance-guided Focused Ultrasound (MRgFUS) is a promising non-invasive thermal ablation technique widely applied in tumor treatment and functional neurological interventions. However, images are often affected by noise and artifacts during treatment, which reduces the accuracy of thermal monitoring and may lead to misjudgments of therapeutic efficacy or incorrect energy delivery. To address this issue, we propose an adaptive phase image denoising method to improve MRgFUS thermometry with a thermal-response Gaussian model. First, the three-component Gaussian modeling method is applied to the temperature profile to construct a thermal response probability density function. Then, by capturing the tail behavior of the temperature response, our method adaptively adjusts phase image denoising strength. This can preserve fine textural details in the thermal maps as much as possible. Experimental results on real MRgFUS datasets demonstrate that the proposed method not only removes noise but also effectively reduces fluctuations and abrupt changes in the temperature profiles improving the reliability of MRgFUS thermal monitoring.

DDPM-EMF: a denoising diffusion probabilistic model-based feature-enhancement fusion network for medical image fusion

Wed, 01 Jan 2025 00:00:00 +0000

Medical image fusion is crucial in clinical applications, combining data from various medical imaging modalities into a single high-quality image to enhance diagnosis. However, existing fusion algorithms suffer from several limitations, including inadequate feature extraction, leading to detail loss, poor inheritance of complementary information between modalities, and insufficient evaluation of color information in color and grayscale fusion tasks. To address these challenges, we propose a novel, to our knowledge, medical image fusion framework based on the denoising diffusion probabilistic model. Our model adopts a two-stage training strategy: feature extraction and image reconstruction. An edge-enhancement dense block is designed to work with a denoising diffusion probabilistic model as a feature extractor, learning and extracting joint features from multimodal medical images to ensure comprehensive feature extraction. To further integrate meaningful information and enhance the visual quality of fused images, we design a feature-enhanced reconstruction network that amplifies features during the reconstruction process. Additionally, we develop distinct joint loss functions based on the strengths and weaknesses of different modalities, ensuring effective retention of complementary information. In the color and grayscale fusion task, we introduce a multi-channel joint learning method to ensure the retention of complementary information and incorporate a color difference formula to evaluate color retention. Experimental results demonstrate that our proposed method significantly outperforms existing state-of-the-art techniques, producing fused images with improved clarity, enhanced detail preservation, and more effective inheritance of complementary information across modalities.

NLMap-ATVR: A novel combination of nonlinear mapping network and adaptive total variation regularization for MRI denoising

Wed, 01 Jan 2025 00:00:00 +0000

Magnetic resonance imaging (MRI) is an advanced imaging technique that is used to aid in medical diagnosis. However, common noises such as Gaussian and Rician noise can blur details and structures, affect contrast and reduce signal-to-noise ratio (SNR), so MRI denoising technique becomes an critical step to get noise-free MRI images. Traditional methods still have limitations in effectively balancing noise removal and the preservation of image details and structural information. To address the challenge, this paper proposes an MRI image denoising model that combines Nonlinear Mapping Network (NLMap) and Attention Mechanism-guided Adaptive Total Variation Regularization (ATVR). The model includes a NLMap-ATVR network, a crafted joint loss function and a Bayesian optimization framework. Firstly, the network uses an encoder-decoder architecture, combined with ATVR to ensure noise removal. Secondly, the joint loss function includes mean square error (MSE) loss, perceptual loss and ATVR loss, which are used to consider pixel-level and feature-level spatial structural errors to preserve details and structures. Thirdly, a Bayesian optimization framework is applied to automatically tune the hyperparameters to obtain optimal parameters. Compared with State-of-the-art methods, both subjective and objective evaluations based on experimental results demonstrate that the proposed method not only effectively removes noise but also significantly preserves details and structural information, which greatly improves SNR.