mrimaster

MRI Deep Resolve

Deep Resolve

Deep Resolve MRI is an advanced medical imaging technology that combines deep learning and artificial intelligence (AI) techniques to enhance the process of magnetic resonance imaging (MRI) image reconstruction. It encompasses several components, including Deep Resolve Gain, Deep Resolve Sharp, Deep Resolve Boost, and Deep Resolve Swift Brain.

Deep Resolve Gain: Deep Resolve Gain is a denoising solution that focuses on reducing image noise in MRI scans. It takes advantage of the fact that noise is not uniformly distributed across the image. This variation can be due to factors such as coil array geometries and parallel imaging reconstruction techniques. Conventional noise filters operate globally on the entire reconstructed image, but Deep Resolve Gain incorporates specific noise maps generated from the original raw data. These noise maps are acquired along with the raw data and are used in the image reconstruction process to perform targeted denoising. This approach allows for stronger denoising in areas where noise is most dominant when compared to traditional methods.

T2 TSE image without deep resolve

deep reslve off

Matrix size 224x224 and NEX 1 scantime : 1.20minute

Deep Resolve Gain

deep resove gain

Matrix size 224x224 and NEX 1 scantime : 1.20minute

Deep Resolve Sharp: Deep Resolve Sharp is an image reconstruction technology designed to enhance image sharpness. It employs a deep neural network trained on a vast dataset of low-resolution and high-resolution image pairs. This network is used to generate a high-resolution image from low-resolution input data. Unlike traditional interpolation methods that expand k-space with zeros, Deep Resolve Sharp uses the neural network to predict information in remote areas within k-space, thereby improving image sharpness. The acquired raw data remains an integral part of the reconstruction process to ensure robust and accurate results.

T2 TSE image without deep resolve

T2 TSE images with out Deep Resolve

Matrix size 192x192 and NEX 1 scantime : 1.30minute

Deep Resolve Gain + Sharp

Deep Resolve Gain + Sharp image

Matrix size 265X256 and NEX 1 scantime : 1.30minute

Deep Resolve Boost: Deep Resolve Boost is a raw data-to-image deep learning reconstruction technology that accelerates image acquisition while maintaining high signal-to-noise (SNR) levels. This technology can be applied throughout the body and can be combined with Siemens Healthineers’ Simultaneous Multi-Slice (SMS) technique to achieve even greater acceleration. Deep Resolve Boost allows for significantly faster scan times without sacrificing image resolution or quality.

T2 TSE image without deep resolve

T2 TSE images with out Deep Resolve 1

Matrix size 192X192 and NEX 1 scantime : 1minute

Deep Resolve Boost

Deep Resolve Boost image

Matrix size 192X192 and NEX 1 scantime : 1minute

Advantages of Deep Resolve MRI:

Improved Image Quality: Deep Resolve Gain reduces noise and enhances image clarity, while Deep Resolve Sharp increases image sharpness, resulting in better diagnostic accuracy.

Faster Scan Times: Deep Resolve Boost accelerates image acquisition without compromising signal-to-noise ratio, reducing patient discomfort and enhancing workflow efficiency.

T2 TSE image without deep resolve

t2 tse image

Matrix size 256X256, NEX 2 and IPAT: OFF | scantime : 2.10minute

Deep Resolve Boost

deep resolve boost image

Matrix size 240X240, NEX 1 and IPAT: GRAPPA 2 | scantime : 1.05minute

Artifacts Associated with Deep Resolve MRI:

Noise Artifacts: While Deep Resolve Gain reduces noise, improper handling of noise maps may lead to residual noise artifacts in the reconstructed images.

Sharpness Artifacts: Deep Resolve Sharp’s focus on increasing sharpness may lead to accentuated high-frequency noise or unrealistic edge enhancements.

Motion Artifacts: Rapid image acquisition enabled by Deep Resolve Boost and Swift Brain may be susceptible to motion artifacts in patients who struggle to remain still.

Aliasing Artifacts: There are situations where the Deep Resolve algorithms may encounter difficulty accurately portraying intricate anatomical formations, potentially resulting in distortions or the emergence of aliasing artifacts. The occurrence of aliasing artifacts tends to be more frequent in areas demanding substantial oversampling, such as in the sagittal view of the spine, for instance.

Deep Resolve Gain Sharpness Artifacts

Deep Resolve Gain Sharpness Artifacts 1

Deep Resolve Gain Sharpness Artifacts

Deep Resolve Gain Sharpness Artifacts

Deep Resolve Boost aliasing Artifacts

Deep Resolve Boost artifact

Deep Resolve Boost enhances motion artifact

Deep Resolve Boost enhances motion artifact

References

  • Behl, N. (2021). Deep Resolve – Mobilizing the Power of Networks. Global Marketing Manager MRI Systems, Siemens Healthineers, Erlangen, Germany. MAGNETOM Flash (78), 1/2021.
  • Lopez Schmidt, I., Haag, N., Shahzadi, I., Frohwein, L. J., Schneider, C., Niehoff, J. H., Kroeger, J. R., Borggrefe, J., Moenninghoff, C. (2023). Diagnostic Image Quality of a Low-Field (0.55T) Knee MRI Protocol Using Deep Learning Image Reconstruction Compared with a Standard (1.5T) Knee MRI Protocol. J Clin Med, 12(5), 1916. doi: 10.3390/jcm12051916.