Overall, AI-assisted MRI reconstruction represents a significant advancement in medical imaging and has considerable potential to improve diagnostic accuracy, clinical workflow, and patient-centred MRI practice by enabling faster, more reliable image reconstruction.
Abstract
Magnetic resonance imaging (MRI) is an essential diagnostic imaging modality that provides excellent soft-tissue contrast without ionizing radiation; however, prolonged acquisition time remains a major limitation that may reduce patient comfort, increase motion artifacts, and restrict clinical workflow. Conventional reconstruction techniques, including Fourier reconstruction, parallel imaging, and compressed sensing, have improved imaging efficiency but remain constrained by noise amplification, reconstruction artifacts, and limited performance at higher acceleration factors. Recent advances in artificial intelligence (AI), particularly deep learning, have transformed MRI image reconstruction by enabling rapid generation of high-quality images from under sampled k-space data. AI-based reconstruction techniques improve image quality through effective noise suppression, artifact reduction, preservation of anatomical details, and shortened reconstruction time. This review summarizes the fundamental principles of MRI image reconstruction, the evolution from conventional reconstruction methods to AI-assisted approaches, and the major deep learning architectures used in contemporary MRI reconstruction, including convolutional neural networks, U-Net, unrolled optimization networks, generative adversarial networks, and transformer-based models. Furthermore, current clinical applications of neuroimaging, musculoskeletal, cardiovascular, abdominal, and oncologic MRI are discussed alongside existing challenges and future research directions. Overall, AI-assisted MRI reconstruction represents a significant advancement in medical imaging and has considerable potential to improve diagnostic accuracy, clinical workflow, and patient-centred MRI practice by enabling faster, more reliable image reconstruction.
Magnetic resonance imaging (MRI) is an essential modality in contemporary medical diagnosis, offering high-resolution images with excellent soft-tissue contrast of internal organs in a non-invasive, non-ionizing, and non-carcinogenic manner. However, the primary drawback of this technique is a lengthy data-acquisition process, which raises the possibility of picking up artefacts as well as discomfort in patients. To accelerate it, data is acquired at sub-Nyquist rates, and the image is recovered from undersampled, noisy measurements using the compressed sensing (CS) technique. Composite priors in CS–MRI have demonstrated superior performance compared to single priors, despite having limited image priors, which also leads to slower reconstruction speed. To date, deep unfolded networks (DUNs), which integrate the powerful data-driven prior of traditional deep learning (DL) with the strong interpretability of optimization algorithms, are the most outstanding reconstruction technology in MRI acceleration, yielding superior and interpretable results. This study proposes the iterative three-operator splitting (TOS) method to solve the composite regularization problem and a composite denoising network (CDNet) that maintains consistent arithmetic structures with it. The CDNet is implemented using complex-valued DL strategies for richer representation of MRI. Extensive experiments using brain and knee raw k-space from the FastMRI dataset with acceleration factors (AF) ranging from ×2 to ×10 and assessed using peak signal–to–noise ratio (PSNR), structural similarity index (SSIM), and normalized root–mean–squared error (NRMSE) validate the CDNet. Across AFs, CDNet achieved dataset-averaged PSNR/SSIM/NRMSE of 26.50–33.47dB/0.7787–0.8992/0.2562–0.1303 for brain MRI and 28.08–35.52dB/0.7069–0.8734/0.1929–0.0950 for knee MRI. The proposed network yields superior reconstruction accuracy and faster reconstruction speed compared to the iterative TOS for composite regularization, demonstrating the effectiveness of deep unrolling in accelerating iterative optimization. Furthermore, CDNet outperforms other state-of-the-art DUNs in reconstruction accuracy at comparable inference speed, demonstrating its vast potential for improving patient care, scanner throughput, and healthcare economics.
David Muigai, E. Mwangi, Henry Kiragu· International Journal of Adv...· 0 citations
Medical imaging is necessary for the diagnosis, treatment, and monitoring of sickness. Technology, background noise, and patient features can all contribute to low contrast in medical images. Diagnostic errors are caused by poor contrast, which obscures important anatomical structures and disease characteristics. To enhance lowcontrast medical images, the study employed deep learning in conjunction with conventional image processing techniques. Combining CNNs with sophisticated image enhancement techniques such as histogram equalization, contrast stretching, and noise reduction, this system produces high-quality results. Automated contrast enhancement with structural preservation is a capability of deep learning models that learn complicated visual cues. Supervised and unsupervised learning lead to better X-ray, CT, and MRI generalization. By incorporating attention processes, the model is able to better strengthen its focus on diagnostically relevant regions. Images are uniformly and artifact-free post-processed by the system. Improvements have been made to the image’s contrast, clarity, and diagnostic accuracy. PSNR and SSIM improve the quality of the image. Clinical decision-making and patient outcomes can be improved using automated, robust medical image enhancement solutions that deep learning and image processing can provide, as demonstrated in this work.
Arivu Selvam, T. Ranjitha, M. Al-Mousa et al.· International Conference on...· 0 citations
I2SB-Inversion is proposed, a multi-contrast guided reconstruction framework based on the Schrödinger Bridge that achieves a a high acceleration factor of R=11.38 and consistently outperforms existing methods in both quantitative and qualitative evaluations.
Yue Wang, Yuanbiao Yang, Zhuo-xu Cui et al.· IEEE Transactions on Medical...· 0 citations
Magnetic resonance imaging (MRI) is a widely used non-invasive medical imaging tool. Recent advancements include whole-body low-field MRI systems (0.55T) to address challenges like high cost and infrastructure demands of high-field MRI. However, low-field images suffer from low signal-to-noise ratio (SNR), which reduces image quality and limits spatial resolution. Although averaging multiple images can achieve the desired SNR while maintaining high spatial resolution, this approach increases total scan time which compromises patient comfort, elevates the risk of motion artifacts, and reduces clinical throughput. Post-processing denoising techniques, such as a patch-based low rank denoiser (PROST) and block match ing 3D filtering (BM3D) have been proposed to address this issue, and more recently, deep learning (DL) methods (both supervised and self-supervised). However, supervised DL requires large datasets, often unavailable for low-field MRI. Most notably, there is no clarify on which of these methods works best, nor a systematic comparison exists for kneeimaging at 0.55T. In this work, we address this by implementing and evaluating two classical methods (PROST, BM3D) and four learning-based approaches (self-supervised: Noisier2Noise,Blind2Unblind and Accelerated Deep Image Prior; supervised: DnCNN) on both public fastMRI (3T/1.5T) and In-house 0.55T knee MRI datasets. Performance is analyzed across configurations to determine optimal denoising strategies and whether self-supervised deep learning-based methods are more effective compared to classical and supervised DL based techniques. Results show that: 1) DnCNN outperforms all methods on the fastMRI database yet it lacks generalization when evaluated on low-field data. 2) On higher simulated levels of noise the self-supervised methods are more effective than classical methods, while on lower levels, classical denoising method are more effective. 3) On 0.55T low-field MRI data, Blind2Unblind is more effective than classical and supervised DL methods. In conclusion, in this study self-supervised methods showed superior performance over supervised and classical methods for low-field knee MR images. However, validation in a larger dataset of knee at 0.55T is needed to further support this conclusion. These findings aim to advance the development of practical denoising strategies that deliver high-quality images while enabling shorter acquisition times and broader accessibility of low-field knee MRI.
Background: CT is a trusted imaging modality for diagnosing disease. The quality of the CT image plays the most important role in disease prediction. Images will be degraded by noise, blur, and other artifacts. Therefore, the enhancement and reconstruction of CT images are required before processing for prediction. Enhancements are only cosmetic changes in images, but reconstruction provides well-defined models for reconstruction. In this paper, we have compared the performance of these four CT brain reconstruction and enhancement techniques: Filtered back projection (FBP), Iterative reconstruction, Generative Adversarial Network (GAN), and CNN autoencoder, by adding Speckle and Poisson noise.
Objectives: The possible objective statements of this study will cover a broad spectrum of developments in medical imaging technology. These are aimed at enhancing diagnostic test quality, patient safety, and overall care outcomes through innovative imaging methods and data analysis. The suggested objectives involve integrating an artificial intelligence algorithm to enhance them, thereby improving patient care and diagnostic accuracy across numerous medical fields. In this experiment, the CT brain stroke dataset was obtained from the Kaggle database and evaluated using performance metrics (PSNR, SSIM, and MSE) across the normal, ischemia, bleeding, and external test folders, with Poisson and Speckle noise added using the GAN, CNN autoencoder, iterative, and FBP techniques.
Materials and methods: The CT brain stroke dataset contains CT images organized in four folders: normal, ischemia, bleeding, and external test. These folders can be used to evaluate performance metrics. The methods applied include: Convolutional Neural Network (CNN) autoencoder, Generative Adversarial Network (GAN), iterative reconstruction, and FBP, each implemented with hyper-tuned parameters. Experimental results calculate PSNR, SSIM, and MSE after adding Poisson and Speckle noise to test the above deep learning algorithms.
Results: The lowest image quality was achieved for FBP for all datasets and both noise types. When the noise was Poisson, the performance of iterative reconstruction was the best, with PSNR exceeding 37 dB, SSIM exceeding 0.972, and MSE as low as 0.0002. On adding Poisson noise, the GAN achieved higher PSNR values and SSIM values up to 32.54 dB and 0.938, respectively, and with low errors in reconstruction, except for the external test folder. The CNN autoencoder (CNN-AE) also yielded significant improvements over FBP, achieving competitive PSNR and SSIM scores across all datasets.
Conclusion: The results indicate that the quality of CT images can be significantly improved using advanced deep learning and Iterative reconstruction methods compared with conventional FBP. It is observed that iterative reconstruction yields the best results for images corrupted by Poisson noise, whereas GANs perform best for images corrupted by Speckle noise. Therefore, reconstruction performance depends on the nature of the noise present in the CT images.
Medical imaging systems such as CT, MRI, PET, and SPECT do not directly acquire images. Instead, they measure physical signals that encode anatomical or physiological information, and image reconstruction recovers the underlying image by solving an inverse problem. Although these imaging modalities are governed by different imaging physics, they share a common computational framework that naturally connects medical physics, linear algebra, probability, numerical optimization, and efficient computing. As medical imaging systems acquire increasingly large and higher-dimensional datasets, image reconstruction has become one of the primary computational bottlenecks in modern medical imaging. Advanced reconstruction methods, including analytical reconstruction, iterative optimization, and statistical model-based reconstruction, substantially improve image quality while reducing radiation dose or scan time, but at significantly increased computational cost. Efficient computing has therefore become essential for achieving clinically practical reconstruction times. This chapter presents a unified computational perspective on medical image acquisition and reconstruction across CT, MRI, PET, and SPECT. It first reviews the imaging physics and data acquisition process for each modality and derives a generalized mathematical framework for image reconstruction. Building on this framework, the chapter discusses analytical, iterative, and statistical reconstruction methods together with their computational characteristics. Finally, it examines efficient computing considerations, including optimization algorithms, physics-aware forward operators, memory-efficient implementations, and parallel computing strategies. Together, these topics demonstrate how the integration of imaging physics, mathematical modeling, and efficient computing enables accurate and scalable medical image reconstruction.
Xiao Wang, Jayasai R. Rajagopal, Md Safaiat Hossain et al.· 0 citations