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.

Anatomically informed multi-level fiber tractography for targeted virtual dissection.

OBJECTIVES: Diffusion-weighted MRI can assist preoperative planning by reconstructing the trajectory of eloquent fiber pathways, such as the corticospinal tract (CST). However, accurate reconstruction of the full extent of the CST remains challenging with existing tractography methods. We suggest a novel tractography algorithm exploiting unused fiber orientations to produce more complete and reliable results. METHODS: Our novel approach, referred to as multi-level fiber tractography (MLFT), reconstructs fiber pathways by progressively considering previously unused fiber orientations at multiple levels of tract propagation. Anatomical priors are used to minimize the number of false-positive pathways. The MLFT method was evaluated on synthetic data and in vivo data by reconstructing the CST while compared to conventional tractography approaches. RESULTS: The radial extent of MLFT reconstructions is comparable to that of probabilistic reconstruction: [Formula: see text] for the left and [Formula: see text] for the right hemisphere according to Wilcoxon test, while achieving significantly higher topography preservation compared to probabilistic tractography: [Formula: see text]. DISCUSSION: MLFT provides a novel way to reconstruct fiber pathways by adding the capability of including branching pathways in fiber tractography. Thanks to its robustness, feasible reconstruction extent and topography preservation, our approach may assist in clinical practice as well as in virtual dissection studies.

Insights from the IronTract challenge: Optimal methods for mapping brain pathways from multi-shell diffusion MRI.

Limitations in the accuracy of brain pathways reconstructed by diffusion MRI (dMRI) tractography have received considerable attention. While the technical advances spearheaded by the Human Connectome Project (HCP) led to significant improvements in dMRI data quality, it remains unclear how these data should be analyzed to maximize tractography accuracy. Over a period of two years, we have engaged the dMRI community in the IronTract Challenge, which aims to answer this question by leveraging a unique dataset. Macaque brains that have received both tracer injections and ex vivo dMRI at high spatial and angular resolution allow a comprehensive, quantitative assessment of tractography accuracy on state-of-the-art dMRI acquisition schemes. We find that, when analysis methods are carefully optimized, the HCP scheme can achieve similar accuracy as a more time-consuming, Cartesian-grid scheme. Importantly, we show that simple pre- and post-processing strategies can improve the accuracy and robustness of many tractography methods. Finally, we find that fiber configurations that go beyond crossing (e.g., fanning, branching) are the most challenging for tractography. The IronTract Challenge remains open and we hope that it can serve as a valuable validation tool for both users and developers of dMRI analysis methods.

Deep learning-based grading of ductal carcinoma in situ in breast histopathology images.

Ductal carcinoma in situ (DCIS) is a non-invasive breast cancer that can progress into invasive ductal carcinoma (IDC). Studies suggest DCIS is often overtreated since a considerable part of DCIS lesions may never progress into IDC. Lower grade lesions have a lower progression speed and risk, possibly allowing treatment de-escalation. However, studies show significant inter-observer variation in DCIS grading. Automated image analysis may provide an objective solution to address high subjectivity of DCIS grading by pathologists. In this study, we developed and evaluated a deep learning-based DCIS grading system. The system was developed using the consensus DCIS grade of three expert observers on a dataset of 1186 DCIS lesions from 59 patients. The inter-observer agreement, measured by quadratic weighted Cohen's kappa, was used to evaluate the system and compare its performance to that of expert observers. We present an analysis of the lesion-level and patient-level inter-observer agreement on an independent test set of 1001 lesions from 50 patients. The deep learning system (dl) achieved on average slightly higher inter-observer agreement to the three observers (o1, o2 and o3) (κo1,dl = 0.81, κo2,dl = 0.53 and κo3,dl = 0.40) than the observers amongst each other (κo1,o2 = 0.58, κo1,o3 = 0.50 and κo2,o3 = 0.42) at the lesion-level. At the patient-level, the deep learning system achieved similar agreement to the observers (κo1,dl = 0.77, κo2,dl = 0.75 and κo3,dl = 0.70) as the observers amongst each other (κo1,o2 = 0.77, κo1,o3 = 0.75 and κo2,o3 = 0.72). The deep learning system better reflected the grading spectrum of DCIS than two of the observers. In conclusion, we developed a deep learning-based DCIS grading system that achieved a performance similar to expert observers. To the best of our knowledge, this is the first automated system for the grading of DCIS that could assist pathologists by providing robust and reproducible second opinions on DCIS grade.

Advanced Ultrasound and Photoacoustic Imaging in Cardiology.

Cardiovascular diseases (CVDs) remain the leading cause of death worldwide. An effective management and treatment of CVDs highly relies on accurate diagnosis of the disease. As the most common imaging technique for clinical diagnosis of the CVDs, US imaging has been intensively explored. Especially with the introduction of deep learning (DL) techniques, US imaging has advanced tremendously in recent years. Photoacoustic imaging (PAI) is one of the most promising new imaging methods in addition to the existing clinical imaging methods. It can characterize different tissue compositions based on optical absorption contrast and thus can assess the functionality of the tissue. This paper reviews some major technological developments in both US (combined with deep learning techniques) and PA imaging in the application of diagnosis of CVDs.

Cross-scanner and cross-protocol multi-shell diffusion MRI data harmonization: Algorithms and results.

Cross-scanner and cross-protocol variability of diffusion magnetic resonance imaging (dMRI) data are known to be major obstacles in multi-site clinical studies since they limit the ability to aggregate dMRI data and derived measures. Computational algorithms that harmonize the data and minimize such variability are critical to reliably combine datasets acquired from different scanners and/or protocols, thus improving the statistical power and sensitivity of multi-site studies. Different computational approaches have been proposed to harmonize diffusion MRI data or remove scanner-specific differences. To date, these methods have mostly been developed for or evaluated on single b-value diffusion MRI data. In this work, we present the evaluation results of 19 algorithms that are developed to harmonize the cross-scanner and cross-protocol variability of multi-shell diffusion MRI using a benchmark database. The proposed algorithms rely on various signal representation approaches and computational tools, such as rotational invariant spherical harmonics, deep neural networks and hybrid biophysical and statistical approaches. The benchmark database consists of data acquired from the same subjects on two scanners with different maximum gradient strength (80 and 300 â€‹mT/m) and with two protocols. We evaluated the performance of these algorithms for mapping multi-shell diffusion MRI data across scanners and across protocols using several state-of-the-art imaging measures. The results show that data harmonization algorithms can reduce the cross-scanner and cross-protocol variabilities to a similar level as scan-rescan variability using the same scanner and protocol. In particular, the LinearRISH algorithm based on adaptive linear mapping of rotational invariant spherical harmonics features yields the lowest variability for our data in predicting the fractional anisotropy (FA), mean diffusivity (MD), mean kurtosis (MK) and the rotationally invariant spherical harmonic (RISH) features. But other algorithms, such as DIAMOND, SHResNet, DIQT, CMResNet show further improvement in harmonizing the return-to-origin probability (RTOP). The performance of different approaches provides useful guidelines on data harmonization in future multi-site studies.

Correcting time-intensity curves in dynamic contrast-enhanced breast MRI for inhomogeneous excitation fields at 7T.

PURPOSE: Inhomogeneous excitation at ultrahigh field strengths (7T and above) compromises the reliability of quantified dynamic contrast-enhanced breast MRI. This can hamper the introduction of ultrahigh field MRI into the clinic. Compensation for this non-uniformity effect can consist of both hardware improvements and post-acquisition corrections. This paper investigated the correctable radiofrequency transmit ( B1+ ) range post-acquisition in both simulations and patient data for 7T MRI. METHODS: Simulations were conducted to determine the minimum B1+ level at which corrections were still beneficial because of noise amplification. Two correction strategies leading to differences in noise amplification were tested. The effect of the corrections on a 7T patient data set (N = 38) with a wide range of B1+ levels was investigated in terms of time-intensity curve types as well as washin, washout and peak enhancement values. RESULTS: In simulations assuming a common amount of T1 saturation, the lowest B1+ level at which the SNR of the corrected images was at least that of the original precontrast image was 43% of the nominal angle. After correction, time-intensity curve types changed in 24% of included patients, and the distribution of curve types corresponded better to the distribution found in literature. Additionally, the overlap between the distributions of washin, washout, and peak enhancement values for grade 1 and grade 2 tumors was slightly reduced. CONCLUSION: Although the correctable range varies with the amount of T1 saturation, post-acquisition correction for inhomogeneous excitation was feasible down to B1+ levels of 43% of the nominal angle in vivo.

Accelerating implant RF safety assessment using a low-rank inverse update method.

PURPOSE: Patients who have medical metallic implants, e.g. orthopaedic implants and pacemakers, often cannot undergo an MRI exam. One of the largest risks is tissue heating due to the radio frequency (RF) fields. The RF safety assessment of implants is computationally demanding. This is due to the large dimensions of the transmit coil compared to the very detailed geometry of an implant. METHODS: In this work, we explore a faster computational method for the RF safety assessment of implants that exploits the small geometry. The method requires the RF field without an implant as a basis and calculates the perturbation that the implant induces. The inputs for this method are the incident fields and a library matrix that contains the RF field response of every edge an implant can occupy. Through a low-rank inverse update, using the Sherman-Woodbury-Morrison matrix identity, the EM response of arbitrary implants can be computed within seconds. We compare the solution from full-wave simulations with the results from the presented method, for two implant geometries. RESULTS: From the comparison, we found that the resulting electric and magnetic fields are numerically equivalent (maximum error of 1.35%). However, the computation was between 171 to 2478 times faster than the corresponding GPU accelerated full-wave simulation. CONCLUSIONS: The presented method enables for rapid and efficient evaluation of the RF fields near implants and might enable situation-specific scanning conditions.

Untangling the diffusion signal using the phasor transform.

Separating the decay signal from diffusion-weighted scans into two or more components can be challenging. The phasor technique is well established in the field of optical microscopy for visualization and separation of fluorescent dyes with different lifetimes. The use of the phasor technique for separation of diffusion-weighted decay signals was recently proposed. In this study, we investigate the added value of this technique for fitting decay models and visualization of decay rates. Phasor visualization was performed in five glioblastoma patients. Using simulations, the influence of incorrect diffusivity values and of the number of b-values on fitting a three-component model with fixed diffusivities (dubbed "unmixing") was investigated for both a phasor-based fit and a linear least squares (LLS) fit. Phasor-based intravoxel incoherent motion (IVIM) fitting was compared with nonlinear least squares (NLLS) and segmented fitting (SF) methods in terms of accuracy and precision. The distributions of the parameter estimates of simulated data were compared with those obtained in a healthy volunteer. In the phasor visualizations of two glioblastoma patients, a cluster of points was observed that was not seen in healthy volunteers. The identified cluster roughly corresponded to the enhanced edge region of the tumor of two glioblastoma patients visible on fluid-attenuated inversion recovery (FLAIR) images. For fitting decay models the usefulness of the phasor transform is less pronounced, but the additional knowledge gained from the geometrical configuration of phasor space can aid fitting routines. This has led to slightly improved fitting results for the IVIM model: phasor-based fitting yielded parameter maps with higher precision than the NLLS and SF methods for parameters f and D (interquartile range [IQR] for f: NLLS 27, SF 12, phasor 5.7%; IQR for D: NLLS 0.28, SF 0.18, phasor 0.10 µm2 /s). For unmixing, LLS fitting slightly but consistently outperformed phasor-based fitting in all of the tested scenarios.

Ultrasound Based Wall Stress Analysis of Abdominal Aortic Aneurysms using Multiperspective Imaging.

BACKGROUND: Current clinical guidelines for surgical repair of abdominal aortic aneurysms (AAAs) are primarily based on maximum diameter assessment. From a biomechanical point of view, not only the diameter but also peak wall stresses will play an important role in rupture risk assessment. These methods require patient specific geometry which typically uses computed tomography (CT) or magnetic resonance imaging. Recently, wall stress analysis based on 3D ultrasound (US) has been proposed, and shows promising results. However, the major limitations in these studies were the use of manual segmentation and the limiting field of view of US. Therefore in this study, the AAA is imaged with multiperspective 3D ultrasound, merged to obtain a large field of view, and afterwards automatically segmented. Geometry and wall stress results were validated using CT imaging. METHODS: Three dimensional US and CT data were available for 40 AAA patients (maximum diameter 34-61 mm). The full US based AAA geometry was determined using automatic segmentation, and when the aneurysm exceeded a single 3D volume, automatic fusion of multiple 3D US volumes was used. Wall stress analysis was performed for all AAA patients and percentile wall stresses were derived. The accuracy of the US based geometry and wall stress prediction was measured by comparison with CT data. RESULTS: Estimated geometries derived from 3D US and CT data showed good similarity, with an overall median similarity index (SI) of 0.89 and interquartile range of 0.87-0.92, whereas the median Hausdorff distances (HD), a measure for the maximum local mismatch, was 4.6 (4.0-5.9) mm for all AAA geometries. Thereby, the wall stress results based on merged multiperspective 3D US data revealed a greater similarity to CT than single 3D US data. CONCLUSION: This study showed that large volume geometry assessment of AAAs using multiperspective 3D ultrasound, segmentation and fusion, and wall stress analysis is feasible in a robust and labour efficient manner.

Reducing distortions in echo-planar breast imaging at ultrahigh field with high-resolution off-resonance maps.

PURPOSE: DWI is a promising modality in breast MRI, but its clinical acceptance is slow. Analysis of DWI is hampered by geometric distortion artifacts, which are caused by off-resonant spins in combination with the low phase-encoding bandwidth of the EPI sequence used. Existing correction methods assume smooth off-resonance fields, which we show to be invalid in the human breast, where high discontinuities arise at tissue interfaces. METHODS: We developed a distortion correction method that incorporates high-resolution off-resonance maps to better solve for severe distortions at tissue interfaces. The method was evaluated quantitatively both ex vivo in a porcine tissue phantom and in vivo in 5 healthy volunteers. The added value of high-resolution off-resonance maps was tested using a Wilcoxon signed rank test comparing the quantitative results obtained with a low-resolution off-resonance map with those obtained with a high-resolution map. RESULTS: Distortion correction using low-resolution off-resonance maps corrected most of the distortions, as expected. Still, all quantitative comparison metrics showed increased conformity between the corrected EPI images and a high-bandwidth reference scan for both the ex vivo and in vivo experiments. All metrics showed a significant improvement when a high-resolution off-resonance map was used (P < 0.05), in particular at tissue boundaries. CONCLUSION: The use of off-resonance maps of a resolution higher than EPI scans significantly improves upon existing distortion correction techniques, specifically by superior correction at glandular tissue boundaries.

Estimating B1+ in the breast at 7 T using a generic template.

Dynamic contrast-enhanced MRI is the workhorse of breast MRI, where the diagnosis of lesions is largely based on the enhancement curve shape. However, this curve shape is biased by RF transmit (B1+ ) field inhomogeneities. B1+ field information is required in order to correct these. The use of a generic, coil-specific B1+ template is proposed and tested. Finite-difference time-domain simulations for B1+ were performed for healthy female volunteers with a wide range of breast anatomies. A generic B1+ template was constructed by averaging simulations based on four volunteers. Three-dimensional B1+ maps were acquired in 15 other volunteers. Root mean square error (RMSE) metrics were calculated between individual simulations and the template, and between individual measurements and the template. The agreement between the proposed template approach and a B1+ mapping method was compared against the agreement between acquisition and reacquisition using the same mapping protocol. RMSE values (% of nominal flip angle) comparing individual simulations with the template were in the range 2.00-4.01%, with mean 2.68%. RMSE values comparing individual measurements with the template were in the range8.1-16%, with mean 11.7%. The agreement between the proposed template approach and a B1+ mapping method was only slightly worse than the agreement between two consecutive acquisitions using the same mapping protocol in one volunteer: the range of agreement increased from ±16% of the nominal angle for repeated measurement to ±22% for the B1+ template. With local RF transmit coils, intersubject differences in B1+ fields of the breast are comparable to the accuracy of B1+ mapping methods, even at 7 T. Consequently, a single generic B1+ template suits subjects over a wide range of breast anatomies, eliminating the need for a time-consuming B1+ mapping protocol.

Arterial and portal venous liver perfusion using selective spin labelling MRI.

PURPOSE: To investigate the feasibility of selective arterial and portal venous liver perfusion imaging with spin labelling (SL) MRI, allowing separate labelling of each blood supply. METHODS: The portal venous perfusion was assessed with a pulsed EPISTAR technique and the arterial perfusion with a pseudo-continuous sequence. To explore precision and reproducibility, portal venous and arterial perfusion were separately quantified in 12 healthy volunteers pre- and postprandially (before and after meal intake). In a subgroup of 6 volunteers, the accuracy of the absolute portal perfusion and its relative postprandial change were compared with MRI flow measurements of the portal vein. RESULTS: The portal venous perfusion significantly increased from 63 ± 22 ml/100g/min preprandially to 132 ± 42 ml/100g/min postprandially. The arterial perfusion was lower with 35 ± 22 preprandially and 22 ± 30 ml/100g/min postprandially. The pre- and postprandial portal perfusion using SL correlated well with flow-based perfusion (r(2) = 0.71). Moreover, postprandial perfusion change correlated well between SL- and flow-based quantification (r(2) = 0.77). The SL results are in range with literature values. CONCLUSION: Selective spin labelling MRI of the portal venous and arterial blood supply successfully quantified liver perfusion. This non-invasive technique provides specific arterial and portal venous perfusion imaging and could benefit clinical settings where contrast agents are contraindicated. KEY POINTS: • Perfusion imaging of the liver by Spin Labelling MRI is feasible • Selective Spin Labelling MRI assessed portal venous and arterial liver perfusion separately • Spin Labelling based portal venous liver perfusion showed significant postprandial increase • Spin Labelling based portal perfusion correlated well with phase-contrast based portal perfusion • This non-invasive technique could benefit settings where contrast agents are contraindicated.

Liver perfusion in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI): comparison of enhancement in Gd-BT-DO3A and Gd-EOB-DTPA in normal liver parenchyma.

PURPOSE: Within-patient comparison of the enhancement patterns of normal liver parenchyma after gadobutrol and gadoxetate disodium, with emphasis on the start of hepatocytic uptake of gadoxetate disodium. MATERIALS AND METHODS: Twenty-one patients (12 female, 9 male) without chronic liver disease underwent 1.5-T contrast-enhanced MRI twice, once with an extracellular contrast agent (gadobutrol) and once with a hepatospecific agent (gadoxetate disodium), using a T1-weighted keyhole sequence. Fifteen whole-liver datasets were acquired up to 5 min for both contrast agents and two additional datasets, up to 20 min, for gadoxetate. Signal intensities (SI) of the parenchyma, aorta and portal vein were measured and analysed relative to pre-contrast parenchymal SI. RESULTS: After gadoxetate, in 29% of the patients the parenchymal SI decreased by ≥5% after the initial vascular-phase-induced peak, while in the other 71% the parenchymal SI remained stable or gradually increased until up to 20 min after the initial peak. The hepatocytic gadoxetate uptake started at a mean of 37.8 s (SD 14.7 s) and not later than 76 s after left ventricle enhancement. CONCLUSION: Parenchymal enhancement due to hepatocytic uptake of gadoxetate can start as early as in the late arterial phase. This may confound the assessment of lesion appearance as compared to extracellular contrast such as gadobutrol. KEY POINTS: Gadoxetate-enhanced liver MRI results in early enhancement of normal parenchyma in patients The start of the hepatobiliary phase coincides with the late arterial phase. This may confound the assessment of lesion appearance compared to extracellular contrast. Different parenchymal enhancement patterns after gadoxetate were found for normal parenchyma.

Influence of learned landmark correspondences on lung CT registration.

BACKGROUND: Disease or injury may cause a change in the biomechanical properties of the lungs, which can alter lung function. Image registration can be used to measure lung ventilation and quantify volume change, which can be a useful diagnostic aid. However, lung registration is a challenging problem because of the variation in deformation along the lungs, sliding motion of the lungs along the ribs, and change in density. PURPOSE: Landmark correspondences have been used to make deformable image registration robust to large displacements. METHODS: To tackle the challenging task of intra-patient lung computed tomography (CT) registration, we extend the landmark correspondence prediction model deep convolutional neural network-Match by introducing a soft mask loss term to encourage landmark correspondences in specific regions and avoid the use of a mask during inference. To produce realistic deformations to train the landmark correspondence model, we use data-driven synthetic transformations. We study the influence of these learned landmark correspondences on lung CT registration by integrating them into intensity-based registration as a distance-based penalty. RESULTS: Our results on the public thoracic CT dataset COPDgene show that using learned landmark correspondences as a soft constraint can reduce median registration error from approximately 5.46 to 4.08 mm compared to standard intensity-based registration, in the absence of lung masks. CONCLUSIONS: We show that using landmark correspondences results in minor improvements in local alignment, while significantly improving global alignment.

Effective deep-learning brain MRI super resolution using simulated training data.

BACKGROUND: In the field of medical imaging, high-resolution (HR) magnetic resonance imaging (MRI) is essential for accurate disease diagnosis and analysis. However, HR imaging is prone to artifacts and is not universally available. Consequently, low-resolution (LR) MRI images are typically acquired. Deep learning (DL)-based super-resolution (SR) techniques can transform LR images into HR quality. However, these techniques require paired HR-LR data for training the SR networks. OBJECTIVE: This research aims to investigate the potential of simulated brain MRI data to train DL-based SR networks. METHODS: We simulated a large set of anatomically diverse, voxel-aligned, and artifact-free brain MRI data at different resolutions. We utilized this simulated data to train four distinct DL-based SR networks and augment their training. The trained networks were then evaluated using real data from various sources. RESULTS: With our trained networks, we produced 0.7mm SR images from standard 1mm resolution multi-source T1w brain MRI. Our experimental results demonstrate that the trained networks significantly enhance the sharpness of LR input MR images. For single-source images, the performance of networks trained solely on simulated data is slightly inferior to those trained solely on real data, with an average structural similarity index (SSIM) difference of 0.025. However, networks augmented with simulated data outperform those trained on single-source real data when evaluated across datasets from multiple sources. CONCLUSION: Paired HR-LR simulated brain MRI data is suitable for training and augmenting diverse brain MRI SR networks. Augmenting the training data with simulated data can enhance the generalizability of the SR networks across real datasets from multiple sources.

Multi-modal brain tumor segmentation via conditional synthesis with Fourier domain adaptation.

Accurate brain tumor segmentation is critical for diagnosis and treatment planning, whereby multi-modal magnetic resonance imaging (MRI) is typically used for analysis. However, obtaining all required sequences and expertly labeled data for training is challenging and can result in decreased quality of segmentation models developed through automated algorithms. In this work, we examine the possibility of employing a conditional generative adversarial network (GAN) approach for synthesizing multi-modal images to train deep learning-based neural networks aimed at high-grade glioma (HGG) segmentation. The proposed GAN is conditioned on auxiliary brain tissue and tumor segmentation masks, allowing us to attain better accuracy and control of tissue appearance during synthesis. To reduce the domain shift between synthetic and real MR images, we additionally adapt the low-frequency Fourier space components of synthetic data, reflecting the style of the image, to those of real data. We demonstrate the impact of Fourier domain adaptation (FDA) on the training of 3D segmentation networks and attain significant improvements in both the segmentation performance and prediction confidence. Similar outcomes are seen when such data is used as a training augmentation alongside the available real images. In fact, experiments on the BraTS2020 dataset reveal that models trained solely with synthetic data exhibit an improvement of up to 4% in Dice score when using FDA, while training with both real and FDA-processed synthetic data through augmentation results in an improvement of up to 5% in Dice compared to using real data alone. This study highlights the importance of considering image frequency in generative approaches for medical image synthesis and offers a promising approach to address data scarcity in medical imaging segmentation.

Brain MR image simulation for deep learning based medical image analysis networks.

BACKGROUND AND OBJECTIVE: As large sets of annotated MRI data are needed for training and validating deep learning based medical image analysis algorithms, the lack of sufficient annotated data is a critical problem. A possible solution is the generation of artificial data by means of physics-based simulations. Existing brain simulation data is limited in terms of anatomical models, tissue classes, fixed tissue characteristics, MR sequences and overall realism. METHODS: We propose a realistic simulation framework by incorporating patient-specific phantoms and Bloch equations-based analytical solutions for fast and accurate MRI simulations. A large number of labels are derived from open-source high-resolution T1w MRI data using a fully automated brain classification tool. The brain labels are taken as ground truth (GT) on which MR images are simulated using our framework. Moreover, we demonstrate that the T1w MR images generated from our framework along with GT annotations can be utilized directly to train a 3D brain segmentation network. To evaluate our model further on larger set of real multi-source MRI data without GT, we compared our model to existing brain segmentation tools, FSL-FAST and SynthSeg. RESULTS: Our framework generates 3D brain MRI for variable anatomy, sequence, contrast, SNR and resolution. The brain segmentation network for WM/GM/CSF trained only on T1w simulated data shows promising results on real MRI data from MRBrainS18 challenge dataset with a Dice scores of 0.818/0.832/0.828. On OASIS data, our model exhibits a close performance to FSL, both qualitatively and quantitatively with a Dice scores of 0.901/0.939/0.937. CONCLUSIONS: Our proposed simulation framework is the initial step towards achieving truly physics-based MRI image generation, providing flexibility to generate large sets of variable MRI data for desired anatomy, sequence, contrast, SNR, and resolution. Furthermore, the generated images can effectively train 3D brain segmentation networks, mitigating the reliance on real 3D annotated data.

Deep-learning-based joint rigid and deformable contour propagation for magnetic resonance imaging-guided prostate radiotherapy.

BACKGROUND: Deep learning-based unsupervised image registration has recently been proposed, promising fast registration. However, it has yet to be adopted in the online adaptive magnetic resonance imaging-guided radiotherapy (MRgRT) workflow. PURPOSE: In this paper, we design an unsupervised, joint rigid, and deformable registration framework for contour propagation in MRgRT of prostate cancer. METHODS: Three-dimensional pelvic T2-weighted MRIs of 143 prostate cancer patients undergoing radiotherapy were collected and divided into 110, 13, and 20 patients for training, validation, and testing. We designed a framework using convolutional neural networks (CNNs) for rigid and deformable registration. We selected the deformable registration network architecture among U-Net, MS-D Net, and LapIRN and optimized the training strategy (end-to-end vs. sequential). The framework was compared against an iterative baseline registration. We evaluated registration accuracy (the Dice and Hausdorff distance of the prostate and bladder contours), structural similarity index, and folding percentage to compare the methods. We also evaluated the framework's robustness to rigid and elastic deformations and bias field perturbations. RESULTS: The end-to-end trained framework comprising LapIRN for the deformable component achieved the best median (interquartile range) prostate and bladder Dice of 0.89 (0.85-0.91) and 0.86 (0.80-0.91), respectively. This accuracy was comparable to the iterative baseline registration: prostate and bladder Dice of 0.91 (0.88-0.93) and 0.86 (0.80-0.92). The best models complete rigid and deformable registration in 0.002 (0.0005) and 0.74 (0.43) s (Nvidia Tesla V100-PCIe 32 GB GPU), respectively. We found that the models are robust to translations up to 52 mm, rotations up to 15 ∘ $^\circ$ , elastic deformations up to 40 mm, and bias fields. CONCLUSIONS: Our proposed unsupervised, deep learning-based registration framework can perform rigid and deformable registration in less than a second with contour propagation accuracy comparable with iterative registration.

LYSTO: The Lymphocyte Assessment Hackathon and Benchmark Dataset.

We introduce LYSTO, the Lymphocyte Assessment Hackathon, which was held in conjunction with the MICCAI 2019 Conference in Shenzhen (China). The competition required participants to automatically assess the number of lymphocytes, in particular T-cells, in images of colon, breast, and prostate cancer stained with CD3 and CD8 immunohistochemistry. Differently from other challenges setup in medical image analysis, LYSTO participants were solely given a few hours to address this problem. In this paper, we describe the goal and the multi-phase organization of the hackathon; we describe the proposed methods and the on-site results. Additionally, we present post-competition results where we show how the presented methods perform on an independent set of lung cancer slides, which was not part of the initial competition, as well as a comparison on lymphocyte assessment between presented methods and a panel of pathologists. We show that some of the participants were capable to achieve pathologist-level performance at lymphocyte assessment. After the hackathon, LYSTO was left as a lightweight plug-and-play benchmark dataset on grand-challenge website, together with an automatic evaluation platform.