Accurate detection of vascular occlusions during endovascular thrombectomy (EVT) is critical in acute ischemic stroke (AIS). Interpretation of digital subtraction angiography (DSA) sequences poses challenges due to anatomical complexity and time constraints. This work proposes OccluNet, a spatio-temporal deep learning model that integrates YOLOX, a single-stage object detector, with transformer-based temporal attention mechanisms to automate occlusion detection in DSA sequences. We compared OccluNet with a YOLOv11 baseline trained on either individual DSA frames or minimum intensity projections. Two spatio-temporal variants were explored for OccluNet: pure temporal attention and divided space-time attention. Evaluation on DSA images from the MR CLEAN Registry revealed the model’s capability to capture temporally consistent features, achieving precision and recall of 89.02 here .

Both inflow and partial volume effects (PVE) are sources of error when measuring the arterial input function (AIF) in DCE-MRI. We proposed a method that estimated the perceived pulse number of spins, and then corrected with these such that both effects were simultaneously compensated. Simulation data demonstrated that the reconstructed AIFs showed only marginal bias. In addition, the algorithm yielded highly correlated reconstructed curves over a wide range of PVEs in clinical data. Our findings show that the PVE can be compensated by considering it as part of the inflow correction as it exhibits similar confounding effects when measuring DCE-AIFs.

White matter hyperintensities are an MRI biomarker of Multiple Sclerosis (MS). However, not all white matter changes are visible on conventional, qualitative MRI. We applied a multi-component MR Fingerprinting protocol to identify potential white matter abnormalities based on increased $$$T_2^*$$$-values. FLAIR and MRF scans were performed in 44 MS patients and 12 healthy control subjects. Significant differences were found in the volume of MRF components with 500ms<$$$T_1, T_2^*$$$<2.5s. This volume correlated moderately with white matter damage on structural MR images. The MRF approach identified larger abnormal tissue volumes than those visible on the structural scans.

Both inflow and the partial volume effect (PVE) are sources of error when measuring the arterial input function (AIF) in dynamic contrast-enhanced (DCE) MRI. This is relevant, as errors in the AIF can propagate into pharmacokinetic parameter estimations from the DCE data. A method was introduced for flow correction by estimating and compensating the number of the perceived pulse of spins during inflow. We hypothesized that the PVE has an impact on concentration-time curves similar to inflow. Therefore, we aimed to study the efficiency of this method to compensate for both effects simultaneously. We first simulated an AIF with different levels of inflow and PVE contamination. The peak, full width at half-maximum (FWHM), and area under curve (AUC) of the reconstructed AIFs were compared with the true (simulated) AIF. In clinical data, the PVE was included in AIFs artificially by averaging the signal in voxels surrounding a manually selected point in an artery. Subsequently, the artificial partial volume AIFs were corrected and compared with the AIF from the selected point. Additionally, corrected AIFs from the internal carotid artery (ICA), the middle cerebral artery (MCA), and the venous output function (VOF) estimated from the superior sagittal sinus (SSS) were compared. As such, we aimed to investigate the effectiveness of the correction method with different levels of inflow and PVE in clinical data. The simulation data demonstrated that the corrected AIFs had only marginal bias in peak value, FWHM, and AUC. Also, the algorithm yielded highly correlated reconstructed curves over increasingly larger neighbourhoods surrounding selected arterial points in clinical data. Furthermore, AIFs measured from the ICA and MCA produced similar peak height and FWHM, whereas a significantly larger peak and lower FWHM was found compared with the VOF. Our findings indicate that the proposed method has high potential to compensate for PVE and inflow simultaneously. The corrected AIFs could thereby provide a stable input source for DCE analysis.

T2-hyperintense lesions are the key imaging marker of multiple sclerosis (MS). Previous studies have shown that the white matter surrounding such lesions is often also affected by MS. Our aim was to develop a new method to visualize and quantify the extent of white matter tissue changes in MS based on relaxometry properties. We applied a fast, multi-parametric quantitative MRI approach and used a multi-component MR Fingerprinting (MC-MRF) analysis. We assessed the differences in the MRF component representing prolongedrelaxation time between patients with MS and controls and studied the relation between this component’s volume and structural white matter damage identified on FLAIR MRI scans in patients with MS. A total of 48 MS patients at two different sites and 12 healthy controls were scanned with FLAIR and MRF-EPI MRI scans. MRF scans were analyzed with a joint-sparsity multi-component analysis to obtain magnetization fraction maps of different components, representing tissues such as myelin water, white matter, gray matter and cerebrospinal fluid. In the MS patients, an additional component was identified with increased transverse relaxation times compared to the white matter, likely representing changes in free water content. Patients with MS had a higher volume of the long- component in the white matter of the brain compared to healthy controls (B (95%-CI) = 0.004 (0.0006-0.008), p=0.02). Furthermore, this MRF component had a moderate correlation (correlation coefficient R 0.47) with visible structural white matter changes on the FLAIR scans. Also, the component was found to be more extensive compared to structural white matter changes in 73% of MS patients. In conclusion, our MRF acquisition and analysis captured white matter tissue changes in MS patients compared to controls. In patients these tissue changes were more extensive compared to visually detectable white matter changes on FLAIR scans. Our method provides a novel way to quantify the extent of white matter changes in MS patients, which is underestimated using only conventional clinical MRI scans.

Researchers Robiël Naziroglu and Frans Vos have developed a method for improving the assessment of MRI images in cases of Crohn's disease. In time, this may well enable more targeted monitoring and treatment of this chronic intestinal inflammation. Naziroglu will take his doctorate on this subject at TU Delft on Monday 21 November. Better targeting Medical imaging techniques have become essential for the assessment of

Preoperative clinical MRI protocols for gliomas, brain tumors with dismal outcomes due to their infiltrative properties, still rely on conventional structural MRI, which does not deliver information on tumor genotype and is limited in the delineation of diffuse gliomas. The GliMR COST action wants to raise awareness about the state of the art of advanced MRI techniques in gliomas and their possible clinical translation. This review describes current methods, limits, and applications of advanced MRI for the preoperative assessment of glioma, summarizing the level of clinical validation of different techniques. In this second part, we review magnetic resonance spectroscopy (MRS), chemical exchange saturation transfer (CEST), susceptibility-weighted imaging (SWI), MRI-PET, MR elastography (MRE), and MR-based radiomics applications. The first part of this review addresses dynamic susceptibility contrast (DSC) and dynamic contrast-enhanced (DCE) MRI, arterial spin labeling (ASL), diffusion-weighted MRI, vessel imaging, and magnetic resonance fingerprinting (MRF). EVIDENCE LEVEL: 3. TECHNICAL EFFICACY: Stage 2.

In MR Fingerprinting acquisitions, the choice of flip angle sequence has a strong effect on the accuracy of the estimated parameter maps. Undersampling artefacts in time-series images are a dominant source of error for standard MRF estimations. We propose to use an undersampling-error model leveraging on perturbation theory, to optimise flip angle patterns taking into account the k-space readout, a realistic ground truth and signal phase. In vivo scans show strong visible reduction in error. Root-mean-square errors in $$$T_1$$$ and $$$T_2$$$ were reduced with at least 6.4%-points and 9%-points respectively, compared to a Cramér-Rao bound optimised and a conventional pattern.

The use of high undersampling factors and short flip angle trains leads to shorter acquisition times in MR Fingerprinting acquisitions. To obtain accurate multi-component estimates from this data advanced reconstructions are required. We study a low-rank ADMM based reconstruction method that adds a multi-component constraint to the inverse reconstruction problem (MC-ADMM). This method is combined with a joint-sparsity constraint yielding higher quality multi-component estimates with k-SPIJN than with previous methods. In simulations we observed increased stability to sequence truncation and in vivo multi-component estimates contained less noise-like effects.

Glioma growth may be quantified with longitudinal image registration. However, the large mass-effects and tissue changes across images suposse an added challenge. Here, we propose a longitudinal, learning-based and groupwise registration method for the accurate and unbiased registration of glioma MRI. We evaluate on a dataset from the Glioma Longitudinal AnalySiS consortium and compare to classical registration methods. We achieve comparable Dice coeffients, with more detailed registrations, while significantly reducing the runtime to under a minute. The proposed methods may serve as an alternative to classical toolboxes, to provide further insight into glioma growth .

The arterial input function (AIF) plays a crucial role in estimating quantitative perfusion properties from dynamic susceptibility contrast (DSC) MRI. An important issue, however, is that measuring the AIF in absolute contrast-agent concentrations is challenging, due to uncertainty in relation to the measured R 2 ∗ -weighted signal, signal depletion at high concentration, and partial-volume effects. A potential solution could be to derive the AIF from separately acquired dynamic contrast enhanced (DCE) MRI data. We aim to compare the AIF determined from DCE MRI with the AIF from DSC MRI, and estimated perfusion coefficients derived from DSC data using a DCE-driven AIF with perfusion coefficients determined using a DSC-based AIF. AIFs were manually selected in branches of the middle cerebral artery (MCA) in both DCE and DSC data in each patient. In addition, a semi-automatic AIF-selection algorithm was applied to the DSC data. The amplitude and full width at half-maximum of the AIFs were compared statistically using the Wilcoxon rank-sum test, applying a 0.05 significance level. Cerebral blood flow (CBF) was derived with different AIF approaches and compared further. The results showed that the AIFs extracted from DSC scans yielded highly variable peaks across arteries within the same patient. The semi-automatic DSC-AIF had significantly narrower width compared with the manual AIFs, and a significantly larger peak than the manual DSC-AIF. Additionally, the DCE-based AIF provided a more stable measurement of relative CBF and absolute CBF values estimated with DCE-AIFs that were compatible with previously reported values. In conclusion, DCE-based AIFs were reproduced significantly better across vessels, showed more realistic profiles, and delivered more stable and reasonable CBF measurements. The DCE-AIF can, therefore, be considered as an alternative AIF source for quantitative perfusion estimations in DSC MRI.

Preoperative clinical magnetic resonance imaging (MRI) protocols for gliomas, brain tumors with dismal outcomes due to their infiltrative properties, still rely on conventional structural MRI, which does not deliver information on tumor genotype and is limited in the delineation of diffuse gliomas. The GliMR COST action wants to raise awareness about the state of the art of advanced MRI techniques in gliomas and their possible clinical translation or lack thereof. This review describes current methods, limits, and applications of advanced MRI for the preoperative assessment of glioma, summarizing the level of clinical validation of different techniques. In this first part, we discuss dynamic susceptibility contrast and dynamic contrast-enhanced MRI, arterial spin labeling, diffusion-weighted MRI, vessel imaging, and magnetic resonance fingerprinting. The second part of this review addresses magnetic resonance spectroscopy, chemical exchange saturation transfer, susceptibility-weighted imaging, MRI-PET, MR elastography, and MR-based radiomics applications. Evidence Level: 3 Technical Efficacy: Stage 2.

PURPOSE:To develop an efficient algorithm for multicomponent MR fingerprinting (MC-MRF) reconstructions directly from highly undersampled data without making prior assumptions about tissue relaxation times and expected number of tissues. METHODS:The proposed method reconstructs MC-MRF maps from highly undersampled data by iteratively applying a joint-sparsity constraint to the estimated tissue components. Intermediate component maps are obtained by a low-rank multicomponent alternating direction method of multipliers (MC-ADMM) including the non-negativity of tissue weights as an extra regularization term. Over iterations, the used dictionary compression is adjusted. The proposed method (k-SPIJN) is compared with a two-step approach in which image reconstruction and multicomponent estimations are performed sequentially and tested in numerical simulations and in vivo by applying different undersampling factors in eight healthy volunteers. In the latter case, fully sampled data serves as the reference. RESULTS:The proposed method shows improved precision and accuracy in simulations compared with a state-of-art sequential approach. Obtained in vivo magnetization fraction maps for different tissue types show reduced systematic errors and reduced noise-like effects. Root mean square errors in estimated magnetization fraction maps significantly reduce from 13.0% ± $$ \pm $$ 5.8% with the conventional, two-step approach to 9.6% ± $$ \pm $$ 3.9% and 9.6% ± $$ \pm $$ 3.2% with the proposed MC-ADMM and k-SPIJN methods, respectively. Mean standard deviation in homogeneous white matter regions reduced significantly from 8.6% to 2.9% (two step vs. k-SPIJN). CONCLUSION:The proposed MC-ADMM and k-SPIJN reconstruction methods estimate MC-MRF maps from highly undersampled data resulting in improved image quality compared with the existing method.

Objective.Machine Learning methods can learn how to reconstruct magnetic resonance images (MRI) and thereby accelerate acquisition, which is of paramount importance to the clinical workflow. Physics-informed networks incorporate the forward model of accelerated MRI reconstruction in the learning process. With increasing network complexity, robustness is not ensured when reconstructing data unseen during training. We aim to embed data consistency (DC) in deep networks while balancing the degree of network complexity. While doing so, we will assess whether either explicit or implicit enforcement of DC in varying network architectures is preferred to optimize performance.Approach.We propose a scheme called Cascades of Independently Recurrent Inference Machines (CIRIM) to assess DC through unrolled optimization. Herein we assess DC both implicitly by gradient descent and explicitly by a designed term. Extensive comparison of the CIRIM to compressed sensing as well as other Machine Learning methods is performed: the End-to-End Variational Network (E2EVN), CascadeNet, KIKINet, LPDNet, RIM, IRIM, and UNet. Models were trained and evaluated on T1-weighted and FLAIR contrast brain data, and T2-weighted knee data. Both 1D and 2D undersampling patterns were evaluated. Robustness was tested by reconstructing 7.5× prospectively undersampled 3D FLAIR MRI data of multiple sclerosis (MS) patients with white matter lesions.Main results.The CIRIM performed best when implicitly enforcing DC, while the E2EVN required an explicit DC formulation. Through its cascades, the CIRIM was able to score higher on structural similarity and PSNR compared to other methods, in particular under heterogeneous imaging conditions. In reconstructing MS patient data, prospectively acquired with a sampling pattern unseen during model training, the CIRIM maintained lesion contrast while efficiently denoising the images.Significance.The CIRIM showed highly promising generalization capabilities maintaining a very fair trade-off between reconstructed image quality and fast reconstruction times, which is crucial in the clinical workflow.

MR fingerprinting (MRF) is a promising method for quantitative characterization of tissues. Often, voxel-wise measurements are made, assuming a single tissue-type per voxel. Alternatively, the Sparsity Promoting Iterative Joint Non-negative least squares Multi-Component MRF method (SPIJN-MRF) facilitates tissue parameter estimation for identified components as well as partial volume segmentations. The aim of this paper was to evaluate the accuracy and repeatability of the SPIJN-MRF parameter estimations and partial volume segmentations. This was done (1) through numerical simulations based on the BrainWeb phantoms and (2) using in vivo acquired MRF data from 5 subjects that were scanned on the same week-day for 8 consecutive weeks. The partial volume segmentations of the SPIJN-MRF method were compared to those obtained by two conventional methods: SPM12 and FSL. SPIJN-MRF showed higher accuracy in simulations in comparison to FSL- and SPM12-based segmentations: Fuzzy Tanimoto Coefficients (FTC) comparing these segmentations and Brainweb references were higher than 0.95 for SPIJN-MRF in all the tissues and between 0.6 and 0.7 for SPM12 and FSL in white and gray matter and between 0.5 and 0.6 in CSF. For the in vivo MRF data, the estimated relaxation times were in line with literature and minimal variation was observed. Furthermore, the coefficient of variation (CoV) for estimated tissue volumes with SPIJN-MRF were 10.5% for the myelin water, 6.0% for the white matter, 5.6% for the gray matter, 4.6% for the CSF and 1.1% for the total brain volume. CoVs for CSF and total brain volume measured on the scanned data for SPIJN-MRF were in line with those obtained with SPM12 and FSL. The CoVs for white and gray matter volumes were distinctively higher for SPIJN-MRF than those measured with SPM12 and FSL. In conclusion, the use of SPIJN-MRF provides accurate and precise tissue relaxation parameter estimations taking into account intrinsic partial volume effects. It facilitates obtaining tissue fraction maps of prevalent tissues including myelin water which can be relevant for evaluating diseases affecting the white matter.

Direct detection of magnetic fields elicited by neuronal activity using Magnetic Resonance Imaging (MRI) has been a long standing research goal, due to its potential to overcome limitations that are inherent to BOLD fMRI. The MRI signal can be sensitized to oscillating magnetic fields using spin-lock preparations. However, the susceptibility of spin-lock sequences to system imperfections has so far hindered their translational potential for in vivo experiments. Moreover, the sensitivity of the neuro-current MRI signal to the phase of neuroelectric oscillations generates high variance time courses that are not suited for analysis with traditional fMRI data processing techniques. In this work we study the impact of various MRI system imperfections on neuro-current MRI in simulations. Furthermore we propose Statistical Variance Mapping (SVarM) as a new data processing technique for generating activity maps from neurocurrent MRI signal variance. Bloch simulations demonstrated substantial variations of signal intensity for a 400 Hz range of off-resonances and a 360 degrees range of neuro-current oscillating phases. SVarM was tested on synthetic neuro-current data simulated with various degrees of system imperfections. The proposed technique was compared to the previously developed NEMO processing, which is based on mean analysis of time courses. Simulation results show improved resilience against B-0 inhomogeneities with SVarM compared with NEMO processing (Dice coefficient of activation maps: 64.07% SVarM, 57.76% NEMO, p<0.02). Comparable or slightly improved robustness against B-1(+) inhomogeneities was observed as well as higher sensitivity in the absence of B-1(+) inhomogeneities (Dice score of activation maps: 58.34% SVarM, 49.70% NEMO, p<0.01). Finally, SVarM achieved better specificity for low SNR, resulting in activation maps with fewer false positive voxels (FP rate: 0.53% SVarM, 19.28% NEMO, p<0.01). These results underscore the importance of dedicated data processing methods and robust pulse sequences to facilitate the widespread use of direct neuro-current MRI in the presence of system imperfections.

Demyelination is the key pathological process in multiple sclerosis (MS). The extent of demyelination can be quantified with magnetic resonance imaging by assessing the myelin water fraction (MWF). However, long computation times and high noise sensitivity hinder the translation of MWF imaging to clinical practice. In this work, we introduce a more efficient and noise robust method to determine the MWF using a joint sparsity constraint and a pre-computed B1+-T2 dictionary.A single component analysis with this dictionary is used in an initial step to obtain a B1+ map. The T2 distribution is then determined from a reduced dictionary corresponding to the estimated B1+ map using a combination of a non-negativity and a joint sparsity constraint.The non-negativity constraint ensures that a feasible solution with non-negative contribution of each T2 component is obtained. The joint sparsity constraint restricts the T2 distribution to a small set of T2 relaxation times shared between all voxels and reduces the noise sensitivity.The applied Sparsity Promoting Iterative Joint NNLS (SPIJN) algorithm can be implemented efficiently, reducing the computation time by a factor of 50 compared to the commonly used regularized non-negative least squares algorithm. The proposed method was validated in simulations and in 8 healthy subjects with a 3D multi-echo gradient- and spin echo scan at 3 ​T. In simulations, the absolute error in the MWF decreased from 0.031 to 0.013 compared to the regularized NNLS algorithm for SNR ​= ​250. The in vivo results were consistent with values reported in literature and improved MWF-quantification was obtained especially in the frontal white matter. The maximum standard deviation in mean MWF in different regions of interest between subjects was smaller for the proposed method (0.0193) compared to the regularized NNLS algorithm (0.0266). In conclusion, the proposed method for MWF estimation is less computationally expensive and less susceptible to noise compared to state of the art methods. These improvements might be an important step towards clinical translation of MWF measurements.

Objective: To allow efficient learning using the Recurrent Inference Machine (RIM) for image reconstruction whereas not being strictly dependent on the training data distribution so that unseen modalities and pathologies are still accurately recovered. Methods: Theoretically, the RIM learns to solve the inverse problem of accelerated-MRI reconstruction whereas being robust to variable imaging conditions. The efficiency and generalization capabilities with different training datasets were studied, as well as recurrent network units with decreasing complexity: the Gated Recurrent Unit (GRU), the Minimal Gated Unit (MGU), and the Independently Recurrent Neural Network (IndRNN), to reduce inference times. Validation was performed against Compressed Sensing (CS) and further assessed based on data unseen during training. A pathology study was conducted by reconstructing simulated white matter lesions and prospectively undersampled data of a Multiple Sclerosis patient. Results: Training on a single modality of 3T T_1-weighted brain data appeared sufficient to also reconstruct 7T T_2^*-weighted brain and 3T T_2-weighted knee data. The IndRNN is an efficient recurrent unit, reducing inference time by 68% compared to CS, whereas maintaining performance. The RIM was able to reconstruct lesions unseen during training more accurately than CS when trained on T_2-weighted knee data. Training on T_1-weighted brain data and on combined data slightly enhanced the signal compared to CS. Conclusion: The RIM is efficient when decreasing its complexity, which reduces the inference time, whereas still being able to reconstruct data and pathology that was unseen during training.

Cribriform growth patterns in prostate carcinoma are associated with poor prognosis. We aimed to introduce a deep learning method to detect such patterns automatically. To do so, convolutional neural network was trained to detect cribriform growth patterns on 128 prostate needle biopsies. Ensemble learning taking into account other tumor growth patterns during training was used to cope with heterogeneous and limited tumor tissue occurrences. ROC and FROC analyses were applied to assess network performance regarding detection of biopsies harboring cribriform growth pattern. The ROC analysis yielded a mean area under the curve up to 0.81. FROC analysis demonstrated a sensitivity of 0.9 for regions larger than 0.0150 mm2 with on average 7.5 false positives. To benchmark method performance for intra-observer annotation variability, false positive and negative detections were re-evaluated by the pathologists. Pathologists considered 9% of the false positive regions as cribriform, and 11% as possibly cribriform; 44% of the false negative regions were not annotated as cribriform. As a final experiment, the network was also applied on a dataset of 60 biopsy regions annotated by 23 pathologists. With the cut-off reaching highest sensitivity, all images annotated as cribriform by at least 7/23 of the pathologists, were all detected as cribriform by the network and 9/60 of the images were detected as cribriform whereas no pathologist labelled them as such. In conclusion, the proposed deep learning method has high sensitivity for detecting cribriform growth patterns at the expense of a limited number of false positives. It can detect cribriform regions that are labelled as such by at least a minority of pathologists. Therefore, it could assist clinical decision making by suggesting suspicious regions.