Accelerating MRI: The Transformative Role of Compressed Sensing in Medical Imaging

Compressed Sensing (CS) is a transformative approach that leverages the sparsity of signals to reduce the number of samples required to reconstruct an image accurately. In the area of Magnetic Resonance Imaging (MRI), CS is particularly valuable, as traditional MRI is time-consuming, often requiring long scan times to capture high-quality images. This can be particularly challenging for patient comfort and for imaging dynamic processes or when fast acquisition is essential, such as in cardiac MRI or when imaging uncooperative patients, like young children or those with claustrophobia.

Fundamentals of MRI and Challenges

MRI is a non-invasive imaging technique that uses a strong magnetic field and radio waves to generate detailed images of the inside of the body. Traditional MRI techniques require sampling the k-space (a frequency space where the MRI data are stored before image reconstruction) at a high rate to satisfy the Nyquist-Shannon sampling theorem, which dictates that the sampling frequency must be at least twice the maximum frequency present in the signal.

However, this process has inherent limitations: it is time-intensive and can lead to patient discomfort and motion artefacts. Reducing scan time is a critical objective, but simply undersampling the k-space leads to aliasing artefacts in the reconstructed images, which are unacceptable in clinical practice.

Compressed Sensing: The Concept

Compressed Sensing emerged as a solution to these challenges by introducing a new data acquisition and reconstruction framework. The fundamental premise of CS is that if a signal is sparse or compressible in some domain (i.e., it contains only a few non-zero coefficients when transformed into an appropriate basis), then it is possible to reconstruct the signal accurately from far fewer samples than what is required by the Nyquist rate.

In MRI, the signal of interest is often sparse in the spatial domain or can be made sparse by transforming it into another domain, such as the wavelet domain. CS exploits this sparsity to enable the reconstruction of MR images from undersampled k-space data, significantly reducing the acquisition time without compromising the image quality.

Key Principles of Compressed Sensing in MRI

  • Sparsity: This is the soul of CS. An image is considered sparse if the number of non-zero pixels is small compared to the total number of pixels. In MRI, many images have a natural sparsity because the region of interest occupies only a fraction of the spatial domain, or the image can be transformed into a domain where it becomes sparse.
  • Transform Encoding: By applying a transform such as the Discrete Wavelet Transform (DWT), the MR image is represented in a form where sparsity is enhanced. The transform encoding supports the separation of significant components of the image from those that are not, allowing for selective sampling of the important parts.
  • Random Undersampling: CS employs a non-uniform random sampling pattern to capture k-space data. This randomness is crucial as it avoids structured errors and makes artefacts appear as noise, which is easier to remove than structured aliasing. This technique is in contrast with the traditional uniform sampling.
  • Nonlinear Reconstruction: Reconstructing an image from undersampled data is an ill-posed problem. CS addresses this by using sophisticated optimisation algorithms that enforce sparsity constraints, such as L1-norm minimisation, which promotes sparsity by penalising the sum of the absolute values of the image coefficients.

Implementing Compressed Sensing in MRI

Implementing CS in MRI involves several steps, starting with the design of an appropriate undersampling scheme for the k-space data. The random undersampling pattern is usually designed to meet specific constraints, such as ensuring a minimum level of data for image reconstruction and avoiding noise amplification.

Once the data are acquired, the reconstruction algorithm plays a critical role. The algorithm incorporates a priori knowledge about the sparsity of the image and leverages optimisation techniques to solve the inverse problem. Iterative algorithms, often based on convex optimisation, are employed to reconstruct the image by finding the sparsest solution that is consistent with the observed undersampled data.

Advantages and Clinical Impact

The most direct benefit of CS in MRI is the significant reduction in scan time. This improvement enhances patient comfort, reduces motion artefacts, and increases throughput in clinical settings. Additionally, it enables the capture of dynamic processes in real-time, which is invaluable for certain types of diagnostic imaging, such as cardiac MRI or functional MRI (fMRI).

Furthermore, by reducing the need for prolonged exposure to strong magnetic fields and radiofrequency pulses, CS can potentially reduce the energy deposition in the tissue, known as the specific absorption rate (SAR), which is a safety concern during MRI scans.

Challenges and Future Directions

While the potential of CS in MRI is enormous, its implementation in clinical practice faces challenges. The most significant is the computational demand of the reconstruction process, which traditionally required extensive processing time. However, advances in computational power and the development of more efficient algorithms are mitigating this issue.

Moreover, developing universally accepted undersampling patterns and reconstruction algorithms that can be standardised across different machines and for various types of scans is a complex task. There’s also a need for extensive validation to ensure that the use of CS does not compromise the diagnostic quality of MRI images.

Another area of active research is the integration of machine learning and artificial intelligence with CS to improve reconstruction quality and efficiency further. Machine learning algorithms can learn patterns in the data to optimise sampling schemes and reconstruction processes, potentially opening new frontiers in rapid and dynamic MRI.


Compressed Sensing represents a paradigm shift in MRI technology. By enabling faster scans without sacrificing image quality, CS enhances the efficiency and comfort of MRI procedures. As research progresses and technology advances, we can anticipate broader adoption of CS in clinical MRI, making it a standard tool for efficient and patient-friendly medical imaging.

You Are Here: Home » compressed sensing
Tags: Functional Magnetic Resonance Imaging, MRI
Open Medscience