Unrolled Optimization via Physics-Assisted Convolutional Neural Network for MR-Based Electrical Properties Tomography: A Numerical Investigation.

Magnetic Resonance imaging based Electrical Properties Tomography (MR-EPT) is a non-invasive technique that measures the electrical properties (EPs) of biological tissues. In this work, we present and numerically investigate the performance of an unrolled, physics-assisted method for 2D MR-EPT reconstructions, where a cascade of Convolutional Neural Networks is used to compute the contrast update. Each network takes in input the EPs and the gradient descent direction (encoding the physics underlying the adopted scattering model) and returns as output the updated contrast function. The network is trained and tested in silico using 2D slices of realistic brain models at 128 MHz. Results show the capability of the proposed procedure to reconstruct EPs maps with quality comparable to that of the popular Contrast Source Inversion-EPT, while significantly reducing the computational time.

Deep learning based correction of RF field induced inhomogeneities for T2w prostate imaging at 7 T.

At ultrahigh field strengths images of the body are hampered by B1 -field inhomogeneities. These present themselves as inhomogeneous signal intensity and contrast, which is regarded as a "bias field" to the ideal image. Current bias field correction methods, such as the N4 algorithm, assume a low frequency bias field, which is not sufficiently valid for T2w images at 7 T. In this work we propose a deep learning based bias field correction method to address this issue for T2w prostate images at 7 T. By combining simulated B1 -field distributions of a multi-transmit setup at 7 T with T2w prostate images at 1.5 T, we generated artificial 7 T images for which the homogeneous counterpart was available. Using these paired data, we trained a neural network to correct the bias field. We predicted either a homogeneous image (t-Image neural network) or the bias field (t-Biasf neural network). In addition, we experimented with the single-channel images of the receive array and the corresponding sum of magnitudes of this array as the input image. Testing was carried out on four datasets: the test split of the synthetic training dataset, volunteer and patient images at 7 T, and patient images at 3 T. For the test split, the performance was evaluated using the structural similarity index measure, Wasserstein distance, and root mean squared error. For all other test data, the features Homogeneity and Energy derived from the gray level co-occurrence matrix (GLCM) were used to quantify the improvement. For each test dataset, the proposed method was compared with the current gold standard: the N4 algorithm. Additionally, a questionnaire was filled out by two clinical experts to assess the homogeneity and contrast preservation of the 7 T datasets. All four proposed neural networks were able to substantially reduce the B1 -field induced inhomogeneities in T2w 7 T prostate images. By visual inspection, the images clearly look more homogeneous, which is confirmed by the increase in Homogeneity and Energy in the GLCM, and the questionnaire scores from two clinical experts. Occasionally, changes in contrast within the prostate were observed, although much less for the t-Biasf network than for the t-Image network. Further, results on the 3 T dataset demonstrate that the proposed learning based approach is on par with the N4 algorithm. The results demonstrate that the trained networks were capable of reducing the B1 -field induced inhomogeneities for prostate imaging at 7 T. The quantitative evaluation showed that all proposed learning based correction techniques outperformed the N4 algorithm. Of the investigated methods, the single-channel t-Biasf neural network proves most reliable for bias field correction.

Uncertainty Assessment for Deep Learning Radiotherapy Applications.

In the last 5 years, deep learning applications for radiotherapy have undergone great development. An advantage of radiotherapy over radiological applications is that data in radiotherapy are well structured, standardized, and annotated. Furthermore, there is much to be gained in automating the current laborious workflows in radiotherapy. After the initial peak in the belief in deep learning, researchers have also identified fundamental weaknesses of deep learning. The basic assumption in deep learning is that the training and test data originate from the same data generating process. This is not always clear-cut in clinical practice, eg, data acquired with 2 different scanners of different vendors might not originate from the same data generating process. Furthermore, it is important to realize residual uncertainties remain even if test data arise from the same data generating process as the training data. As deep learning applications are being introduced in clinical radiotherapy workflows, a deep learning model must express to a user when a prediction exceeds a certain uncertainty threshold. The literature on uncertainty assessment for deep learning applications in radiotherapy is still in its infancy; however, quite a body of literature exists on the validity and uncertainty of deep learning models for computer vision applications. This paper tries to explain these general concepts to the radiotherapy community. Concepts of epistemic and aleatoric uncertainties and techniques to model them in deep learning are described in detail. It is discussed how they can be applied to maximize confidence in automated deep learning-driven workflows. Their usage is demonstrated in 3 examples from radiotherapy literature on deep learning applications, ie, dose prediction, synthetic CT generation, and contouring. In the final part, some of the key elements to ensure confidence and automatic alerting that are still missing are discussed. State-of-the-art automatic solutions for checking within-distribution vs out-of-distribution test samples are discussed. However, these methodologies are still immature, and strict QA protocols and close human supervision will still be needed. Nevertheless, deep learning models offer already much value for radiotherapy.

A local multi-transmit coil combined with a high-density receive array for cerebellar fMRI at 7 T.

The human cerebellum is involved in a wide array of functions, ranging from motor control to cognitive control, and as such is of great neuroscientific interest. However, its function is underexplored in vivo, due to its small size, its dense structure and its placement at the bottom of the brain, where transmit and receive fields are suboptimal. In this study, we combined two dense coil arrays of 16 small surface receive elements each with a transmit array of three antenna elements to improve BOLD sensitivity in the human cerebellum at 7 T. Our results showed improved B1+ and SNR close to the surface as well as g-factor gains compared with a commercial coil designed for whole-head imaging. This resulted in improved signal stability and large gains in the spatial extent of the activation close to the surface (<3.5 cm), while good performance was retained deeper in the cerebellum. Modulating the phase of the transmit elements of the head coil to constructively interfere in the cerebellum improved the B1+ , resulting in a temporal SNR gain. Overall, our results show that a dedicated transmit array along with the SNR gains of surface coil arrays can improve cerebellar imaging, at the cost of a decreased field of view and increased signal inhomogeneity.

SAR and temperature distributions in a database of realistic human models for 7 T cardiac imaging.

PURPOSE: To investigate inter-subject variability of B1+ , SAR and temperature rise in a database of human models using a local transmit array for 7 T cardiac imaging. METHODS: Dixon images were acquired of 14 subjects and segmented in dielectric models with an eight-channel local transmit array positioned around the torso for cardiac imaging. EM simulations were done to calculate SAR distributions. Based on the SAR distributions, temperature simulations were performed for exposure times of 6 min and 30 min. Peak local SAR and temperature rise levels were calculated for different RF shim settings. A statistical analysis of the resulting peak local SAR and temperature rise levels was performed to arrive at safe power limits. RESULTS: For RF shim vectors with random phase and uniformly distributed power, a safe average power limit of 35.7 W was determined (first level controlled mode). When RF amplitude and phase shimming was performed on the heart, a safe average power limit of 35.0 W was found. According to Pennes' model, our numerical study suggests a very low probability of exceeding the absolute local temperature limit of 40 °C for a total exposure time of 6 min and a peak local SAR of 20 W/kg. For a 30 min exposure time at 20 W/kg, it was shown that the absolute temperature limit can be exceeded in the case where perfusion does not change with temperature. CONCLUSION: Safe power constraints were found for 7 T cardiac imaging with an eight-channel local transmit array, while considering the inter-subject variability of B1+ , SAR and temperature rise.

Opening a new window on MR-based Electrical Properties Tomography with deep learning.

In the radiofrequency (RF) range, the electrical properties of tissues (EPs: conductivity and permittivity) are modulated by the ionic and water content, which change for pathological conditions. Information on tissues EPs can be used e.g. in oncology as a biomarker. The inability of MR-Electrical Properties Tomography techniques (MR-EPT) to accurately reconstruct tissue EPs by relating MR measurements of the transmit RF field to the EPs limits their clinical applicability. Instead of employing electromagnetic models posing strict requirements on the measured MRI quantities, we propose a data driven approach where the electrical properties reconstruction problem can be casted as a supervised deep learning task (DL-EPT). DL-EPT reconstructions for simulations and MR measurements at 3 Tesla on phantoms and human brains using a conditional generative adversarial network demonstrate high quality EPs reconstructions and greatly improved precision compared to conventional MR-EPT. The supervised learning approach leverages the strength of electromagnetic simulations, allowing circumvention of inaccessible MR electromagnetic quantities. Since DL-EPT is more noise-robust than MR-EPT, the requirements for MR acquisitions can be relaxed. This could be a major step forward to turn electrical properties tomography into a reliable biomarker where pathological conditions can be revealed and characterized by abnormalities in tissue electrical properties.

Intersubject specific absorption rate variability analysis through construction of 23 realistic body models for prostate imaging at 7T.

PURPOSE: For ultrahigh field (UHF) MRI, the expected local specific absorption rate (SAR) distribution is usually calculated by numerical simulations using a limited number of generic body models and adding a safety margin to take into account intersubject variability. Assessment of this variability with a large model database would be desirable. In this study, a procedure to create such a database with accurate subject-specific models is presented. Using 23 models, intersubject variability is investigated for prostate imaging at 7T with an 8-channel fractionated dipole antenna array with 16 receive loops. METHOD: From Dixon images of a volunteer acquired at 1.5T with a mockup array in place, an accurate dielectric model is built. Following this procedure, 23 subject-specific models for local SAR assessment at 7T were created enabling an extensive analysis of the intersubject B1+ and peak local SAR variability. RESULTS: For the investigated setup, the maximum possible peak local SAR ranges from 2.6 to 4.6 W/kg for 8 × 1 W input power. The expected peak local SAR values represent a Gaussian distribution (µ/σ=2.29/0.29 W/kg) with realistic prostate-shimmed phase settings and a gamma distribution Γ(24,0.09) with multidimensional radiofrequency pulses. Prostate-shimmed phase settings are similar for all models. Using 1 generic phase setting, average B1+ reduction is 7%. Using only 1 model, the required safety margin for intersubject variability is 1.6 to 1.8. CONCLUSION: The presented procedure allows for the creation of a customized model database. The results provide valuable insights into B1+ and local SAR variability. Recommended power thresholds per channel are 3.1 W with phase shimming on prostate or 2.6 W for multidimensional pulses.