Carlos Castillo Passi

All Publications

• KomaMRI.jl: An open-source framework for general MRI simulations with GPU acceleration. Magnetic resonance in medicine Castillo-Passi, C., Coronado, R., Varela-Mattatall, G., Alberola-López, C., Botnar, R., Irarrazaval, P. 2023; 90 (1): 329-342

  Abstract

  To develop an open-source, high-performance, easy-to-use, extensible, cross-platform, and general MRI simulation framework (Koma).Koma was developed using the Julia programming language. Like other MRI simulators, it solves the Bloch equations with CPU and GPU parallelization. The inputs are the scanner parameters, the phantom, and the pulse sequence that is Pulseq-compatible. The raw data is stored in the ISMRMRD format. For the reconstruction, MRIReco.jl is used. A graphical user interface utilizing web technologies was also designed. Two types of experiments were performed: one to compare the quality of the results and the execution speed, and the second to compare its usability. Finally, the use of Koma in quantitative imaging was demonstrated by simulating Magnetic Resonance Fingerprinting (MRF) acquisitions.Koma was compared to two well-known open-source MRI simulators, JEMRIS and MRiLab. Highly accurate results (with mean absolute differences below 0.1% compared to JEMRIS) and better GPU performance than MRiLab were demonstrated. In an experiment with students, Koma was proved to be easy to use, eight times faster on personal computers than JEMRIS, and 65% of test subjects recommended it. The potential for designing acquisition and reconstruction techniques was also shown through the simulation of MRF acquisitions, with conclusions that agree with the literature.Koma's speed and flexibility have the potential to make simulations more accessible for education and research. Koma is expected to be used for designing and testing novel pulse sequences before implementing them in the scanner with Pulseq files, and for creating synthetic data to train machine learning models.

  View details for DOI 10.1002/mrm.29635

  View details for PubMedID 36877139

  View details for PubMedCentralID PMC10952765

• 3D Whole-Heart Joint T₁/T_1ρ Mapping and Water-Fat Imaging on a Clinical 0.55-T Low-Field Scanner NMR IN BIOMEDICINE Crabb, M. G., Kunze, K. P., Castillo-Passi, C., Si, D., Littlewood, S. J., Prieto, C., Botnar, R. M. 2026; 39 (2): e70195

  Abstract

  Myocardial maps are conventionally acquired in 2D breath-hold single-parameter scans that are slow and have limited heart coverage. To overcome limitations associated with 2D breath-hold mapping sequences, we develop a novel free-breathing 3D joint T 1 $$ {T}_1 $$ / T 1 ρ $$ {T}_{1\rho } $$ mapping sequence with Dixon encoding to provide co-registered 3D T 1 $$ {T}_1 $$ and T 1 ρ $$ {T}_{1\rho } $$ maps and water-fat volumes with isotropic spatial resolution in a single scan for comprehensive contrast-agent free myocardial tissue characterization and visualisation of the whole-heart anatomy on a clinical 0.55-T MR scanner. The proposed sequence acquires four interleaved 3D volumes with preparation modules to provide T 1 $$ {T}_1 $$ and T 1 ρ $$ {T}_{1\rho } $$ encoding, with data acquired with a two-echo Dixon readout and 2D image navigators to enable 100 % $$ 100\% $$ respiratory scan efficiency. Images were reconstructed with nonrigid respiratory motion-corrected iterative SENSE with multi-dimensional low-rank patch-based denoising, and maps generated by matching with simulated dictionaries. The proposed sequence was tested in phantoms, 11 healthy subjects and 1 patient, and compared with conventional techniques. For phantoms, the proposed 3D T 1 $$ {T}_1 $$ and T 1 ρ $$ {T}_{1\rho } $$ measurements showed good correlation with 2D spin-echo reference measurements. For healthy subjects, septal myocardial tissue mapping values were T 1 = 743 ± 19 ms $$ {T}_1=743\pm 19\kern0.3em \mathrm{ms} $$ and T 1 ρ = 46 . 9 ± 2 . 7 ms $$ {T}_{1\rho }=46.9\pm 2.7\kern0.3em \mathrm{ms} $$ for the proposed sequence, against T 1 = 681 ± 23 ms $$ {T}_1=681\pm 23\kern0.3em \mathrm{ms} $$ and T 1 ρ = 57 . 9 ± 3 . 6 ms $$ {T}_{1\rho }=57.9\pm 3.6\kern0.3em \mathrm{ms} $$ for 2D modified Look-Locker inversion recovery and 2D T 1 ρ $$ {T}_{1\rho } $$ respectively. Promising results were obtained when the proposed mapping was compared to 2D late-gadolinium enhancement imaging in a patient. The proposed approach enables simultaneous 3D whole-heart joint T 1 $$ {T}_1 $$ / T 1 ρ $$ {T}_{1\rho } $$ mapping and water-fat imaging at 0.55 T in a single scan of ≈ 11 $$ \approx 11 $$ min, demonstrating good agreement with conventional techniques in phantoms and healthy subjects, and promising results in a patient.

  View details for DOI 10.1002/nbm.70195

  View details for Web of Science ID 001664379200002

  View details for PubMedID 41431231

  View details for PubMedCentralID PMC12723209

• Versatile and Highly Efficient MRI Simulation of Arbitrary Motion in KomaMRI. Magnetic resonance in medicine Villacorta-Aylagas, P., Castillo-Passi, C. A., Kierulf, R. A., Menchón-Lara, R. M., Rodríguez-Galván, J. R., Sierra-Pallares, J. B., Irarrazaval, P., Alberola-López, C. 2025

  Abstract

  To extend the KomaMRI simulator with motion capabilities that enable the simulation of both simple, parametrizable motion patterns and arbitrarily complex spin trajectories. Additionally, we introduce a novel file format for storing and sharing dynamic digital phantoms.The existing Phantom structure in KomaMRI has been extended with a new motion field, which preserves support for static phantoms and introduces the capability to define dynamic models, composed of either a single or an arbitrary set of elementary motions. Each motion entry is composed of an action, a time curve, and a spin span, allowing for modular definition and full parameter independence across motion components. An HDF5-based phantom file format has been defined, along with two I/O functions within the simulator. Motion-related MRI experiments have been conducted for both illustrative and comparative evaluation.Illustrative experiments reveal motion-related effects commonly observed in real MRI, such as time of flight and phase contrast. Comparative evaluations show strong qualitative and quantitative agreement with results reported by other simulation tools and with at least a three-fold reduction in computation time.The proposed extension adds flexible motion modeling and simulation capabilities to KomaMRI. The proposed new functions allow for accurate and high temporal and spatial resolution, only limited by computational cost.

  View details for DOI 10.1002/mrm.70145

  View details for PubMedID 41145960

• Cardiac Magnetic Resonance Fingerprinting for Simultaneous T1, T2, and Fat-Fraction Quantification at 0.55 T NMR IN BIOMEDICINE Pedraza, D., Castillo-Passi, C., Kunze, K., Botnar, R. M., Prieto, C. 2025; 38 (10): e70143

  Abstract

  Cardiac magnetic resonance fingerprinting (cMRF) has been shown to allow for simultaneous quantitative characterization of myocardial tissue in a single scan. While cMRF has been assessed at 1.5 T and 3 T, its application at 0.55 T has not been demonstrated yet. This study introduces an adapted version of a previously implemented Dixon cMRF sequence designed for simultaneous quantification of T1, T2, and fat fraction (FF) at 1.5 T, to be employed at 0.55 T within a single breath-hold scan. The sequence was developed using the Pulseq environment and employs a radial tiny golden angle acquisition with bipolar readout. Reconstruction was performed using low-rank inversion in combination with a high-dimensional patch-based regularization. The Dixon cMRF technique at 0.55 T was tested on standardized phantoms and 15 healthy volunteers (HVs). cMRF at 0.55 T was compared to spin-echo (SE) and proton density references from phantoms, as well as conventional T1, T2, and FF mapping sequences at 0.55 T. Intrasession and intersession variability was assessed in phantoms and a representative HV. Results showed a good correlation between the proposed cMRF T1, T2, and FF at 0.55 T and the phantom IR-SE references (R2 ≥ 0.98 for T1 and T2, R2 ≥ 0.97 for FF). Intrasession variability was low (8.9 ± 13.8 ms for T1, 0.1 ± 1 ms for T2, and 0.02 ± 0.03% for FF), as was intersession variability (8.2 ± 8.5 ms, 0.4 ± 1.1 ms, and 0.02 ± 0.25%, respectively). In vivo assessments yielded good map quality, with mean myocardial values of 714 ± 24 ms for T1, 49 ± 5.9 ms for T2, and 2.6 ± 0.9% for FF in comparison to 672 ± 40 for T1-MOLLI, 60 ± 5.4 for T2prep-bSSFP, and 4.7 ± 2.4% for 2-echo PDFF, respectively. The technique demonstrated good agreement for T1 and FF, but T2 was underestimated, which is consistent with findings at higher field strengths. Further investigation in a larger cohort of healthy subjects and in patients with cardiovascular disease is warranted.

  View details for DOI 10.1002/nbm.70143

  View details for Web of Science ID 001575798300021

  View details for PubMedID 40944613

• mtrk-A flexible environment for developing open-source MRI pulse sequences MAGNETIC RESONANCE IN MEDICINE Artiges, A., Saimbhi, A., Castillo-Passi, C., Lattanzi, R., Block, K. 2025

  Abstract

  To introduce mtrk, a new open-source tool based on modern software-engineering principles that simplifies pulse-sequence design, implementation, and dissemination.The mtrk framework is vendor-agnostic and relies on a compact and human-readable descriptive language. Users can design pulse sequences using either a Python-based programming interface or an intuitive graphical interface. The graphical interface also allows for visualizing pulse-sequence diagrams. A driver sequence was developed to run mtrk sequences on MR scanners. A spin-echo sequence was designed with mtrk and converted to Pulseq for comparison. Both versions were compared to an equivalent vendor sequence in phantom and in vivo experiments.Images from the mtrk and Pulseq versions were nearly identical and showed over 90% similarity compared to the vendor sequence, despite minor unavoidable design differences. Phantom images matched corresponding synthetic images simulated using the same pulse sequences.The mtrk framework simplifies the development of pulse sequences by providing an intuitive descriptive language and compatibility with the Pulseq format. Users can design and simulate pulse sequences using the graphical interface without any programming experience.

  View details for DOI 10.1002/mrm.70067

  View details for Web of Science ID 001561662000001

  View details for PubMedID 40891382

• Simultaneous liver T1, T2, and ADC MR fingerprinting using optimized motion-compensated diffusion preparations: An initial validation on volunteers. Magnetic resonance in medicine Velasco, C., Castillo-Passi, C., Chaher, N., Karampinos, D. C., Irarrazaval, P., Phinikaridou, A., Botnar, R. M., Prieto, C. 2025

  Abstract

  To develop a novel MR fingerprinting sequence using optimized motion-compensated diffusion preparations for simultaneous T1, T2, and ADC quantification of liver tissue in a single breath-held scan.A radial spoiled gradient echo acquisition with magnetization preparation modules for T1, T2, and ADC encoding is proposed. To compensate for the signal voids generated by the diffusion preparation, the combination of (1) a breath-held scan, (2) peripheral pulse signal triggering, and (3) an optimized motion-compensated diffusion-preparation module is employed. Phantom experiments were performed to test the accuracy of the technique. The sequence was evaluated in 11 healthy subjects in comparison to conventional mapping techniques. Additional in vivo repeatability assessment experiments were performed.T1, T2, and ADC quantification showed good correlation (r2 > 0.9 for all cases) with reference maps in phantoms and good agreement in vivo against clinical scans (bias not significantly different from zero). A peripheral pulse trigger delay of 200 ms was used to reduce cardiovascular motion artifacts. The repeatability tests prove a low interscan coefficient of variation and a high intraclass correlation coefficient of greater than 0.9 for all cases.Simultaneous quantification of T1, T2, and ADC in liver tissue in a single MR fingerprinting scan of ˜16 s has been proposed, enabling a comprehensive evaluation of hepatic disease through co-registered multiparametric imaging. Further studies are warranted to test this approach in patients with suspected diffuse liver disease to evaluate its potential for liver tissue characterization and tumor staging.

  View details for DOI 10.1002/mrm.30622

  View details for PubMedID 40632800

• Simultaneous 3D aortic lumen and vessel wall imaging at 0.55 T at either systole or diastole MAGNETIC RESONANCE IN MEDICINE Paredes, M., Castillo-Passi, C., Kunze, K. P., Fotaki, A., Littlewood, S., Botnar, R. M., Prieto, C. 2025

  Abstract

  To evaluate the feasibility of a novel, non-contrast enhanced, 3D, simultaneous bright-blood, and black-blood sequence (iT2prep-BOOST) for aortic imaging at 0.55 T at either systole or diastole.Simultaneous contrast-free 3D aortic lumen and vessel wall imaging at 0.55 T is achieved using the recently introduced iT2prep-BOOST framework that interleaves the acquisition of two bright blood images (with inversion recovery T2 preparation [T2prep-IR] and no preparation). To enable either systolic or diastolic aortic imaging, three T2 preparation pulses were investigated-an adiabatic RF pulse and two Malcolm-Levitt (MLEV) pulses (MLEV4 and MLEV8)-to improve image quality in regions with high flow and susceptibility. The proposed approach was evaluated in phantom, 10 healthy subjects and 3 patients with suspected cardiovascular disease. Bright- and black-blood images resulting from the three different T2 preparation pulses were compared both qualitatively and quantitatively, using a 4-point Likert scale for vessel sharpness and presence of blood artifacts. Additionally, the contrast ratio between the lumen and myocardium was computed. Aortic measurements, including the aortic annulus area at systole and diastole, cusp-commissure measurement at the aortic root level during diastole, and aortic diameter at the ascending aortic level during diastole were also performed.Excellent or good image quality scores were obtained for both bright- and black-blood images with iT2prep-BOOST at 0.55 T with all three preparation pulses. The use of MLEV8 T2 preparation scheme improves systolic image quality, reducing the presence of artifacts with a significant difference (p < 0.05) at the mid descending aorta level. This scheme also increases the contrast ratio between aortic lumen and myocardium, compared to the previously used adiabatic RF T2 preparation. The aortic root diameter and area were consistent with values reported in the literature for healthy subjects at 1.5 T.The feasibility of a novel, non-contrast-enhanced, 3D aortic imaging framework for simultaneous bright-blood and black-blood imaging was demonstrated at 0.55 T for either systole or diastole, with a scan time of 7 min. Good image quality scores and aortic measurements in agreement with literature values at 1.5 T were achieved with the MLEV8 T2 preparation. Studies in a larger cohort of healthy subjects and patients with aortopathies are warranted.

  View details for DOI 10.1002/mrm.30611

  View details for Web of Science ID 001513255400001

  View details for PubMedID 40548843

• Highly efficient image navigator based 3D whole-heart cardiac MRA at 0.55T MAGNETIC RESONANCE IN MEDICINE Castillo-Passi, C., Kunze, K. P., Crabb, M. G., Munoz, C., Fotaki, A., Neji, R., Irarrazaval, P., Prieto, C., Botnar, R. M. 2025; 93 (2): 689-698

  Abstract

  To develop and evaluate a highly efficient free-breathing and contrast-agent-free three-dimensional (3D) whole-heart Cardiac Magnetic Resonance Angiography (CMRA) sequence at 0.55T.Free-breathing whole-heart CMRA has been previously proposed at 1.5 and 3T. Direct application of this sequence to 0.55T is not possible due to changes in the magnetic properties of the tissues. To enable free-breathing CMRA at 0.55T, pulse sequence design and acquisition parameters of a previously proposed whole-heart CMRA framework are optimized via Bloch simulations. Image navigators (iNAVs) are used to enable nonrigid respiratory motion-correction and 100% respiratory scan efficiency. Patch-based low-rank denoising is employed to accelerate the scan and account for the reduced signal-to-noise ratio at 0.55T. The proposed approach was evaluated on 11 healthy subjects. Image quality was assessed by a clinical expert (1: poor to 5: excellent) for all intrapericardiac structures. Quantitative evaluation was performed by assessing the vessel sharpness of the proximal right coronary artery (RCA).Optimization resulted in an imaging flip angle of 11 0 ∘ $$ 11{0}^{\circ } $$ , fat saturation flip angle of 18 0 ∘ $$ 18{0}^{\circ } $$ , and six k-space lines for iNAV encoding. The relevant cardiac structures and main coronary arteries were visible in all subjects, with excellent image quality (mean 4 . 9 / 5 . 0 $$ 4.9/5.0 $$ ) and minimal artifacts (mean 4 . 9 / 5 . 0 $$ 4.9/5.0 $$ ), with RCA vessel sharpness ( 50 . 3 % ± 9 . 8 % $$ 50.3\%\pm 9.8\% $$ ) comparable to previous studies at 1.5T.The proposed approach enables 3D whole-heart CMRA at 0.55T in a 6-min scan ( 5 . 9 ± 0 . 7   min $$ 5.9\pm 0.7\;\min $$ ), providing excellent image quality, minimal artifacts, and comparable vessel sharpness to previous 1.5T studies. Future work will include the evaluation of the proposed approach in patients with cardiovascular disease.

  View details for DOI 10.1002/mrm.30316

  View details for Web of Science ID 001412668100017

  View details for PubMedID 39415543

  View details for PubMedCentralID PMC11604836

• DeepSPIO: Super Paramagnetic Iron Oxide Particle Quantification Using Deep Learning in Magnetic Resonance Imaging. IEEE transactions on pattern analysis and machine intelligence Maggiora, G. D., Castillo-Passi, C., Qiu, W., Liu, S., Milovic, C., Sekino, M., Tejos, C., Uribe, S., Irarrazaval, P. 2022; 44 (1): 143-153

  Abstract

  The susceptibility of super paramagnetic iron oxide (SPIO) particles makes them a useful contrast agent for different purposes in MRI. These particles are typically quantified with relaxometry or by measuring the inhomogeneities they produced. These methods rely on the phase, which is unreliable for high concentrations. We present in this study a novel Deep Learning method to quantify the SPIO concentration distribution. We acquired the data with a new sequence called View Line in which the field map information is encoded in the geometry of the image. The novelty of our network is that it uses residual blocks as the bottleneck and multiple decoders to improve the gradient flow in the network. Each decoder predicts a different part of the wavelet decomposition of the concentration map. This decomposition improves the estimation of the concentration, and also it accelerates the convergence of the model. We tested our SPIO concentration reconstruction technique with simulated images and data from actual scans from phantoms. The simulations were done using images from the IXI dataset, and the phantoms consisted of plastic cylinders containing agar with SPIO particles at different concentrations. In both experiments, the model was able to quantify the distribution accurately.

  View details for DOI 10.1109/TPAMI.2020.3012103

  View details for PubMedID 32750834

• A Spatial Off-Resonance Correction in Spirals for Magnetic Resonance Fingerprinting. IEEE transactions on medical imaging Coronado, R., Cruz, G., Castillo-Passi, C., Tejos, C., Uribe, S., Prieto, C., Irarrazaval, P. 2021; 40 (12): 3832-3842

  Abstract

  In MR Fingerprinting (MRF), balanced Steady-State Free Precession (bSSFP) has advantages over unbalanced SSFP because it retains the spin history achieving a higher signal-to-noise ratio (SNR) and scan efficiency. However, bSSFP-MRF is not frequently used because it is sensitive to off-resonance, producing artifacts and blurring, and affecting the parametric map quality. Here we propose a novel Spatial Off-resonance Correction (SOC) approach for reducing these artifacts in bSSFP-MRF with spiral trajectories. SOC-MRF uses each pixel's Point Spread Function to create system matrices that encode both off-resonance and gridding effects. We iteratively compute the inverse of these matrices to reduce the artifacts. We evaluated the proposed method using brain simulations and actual MRF acquisitions of a standardized T1/T2 phantom and five healthy subjects. The results show that the off-resonance distortions in T1/T2 maps were considerably reduced using SOC-MRF. For T2, the Normalized Root Mean Square Error (NRMSE) was reduced from 17.3 to 8.3% (simulations) and from 35.1 to 14.9% (phantom). For T1, the NRMS was reduced from 14.7 to 7.7% (simulations) and from 17.7 to 6.7% (phantom). For in-vivo, the mean and standard deviation in different ROI in white and gray matter were significantly improved. For example, SOC-MRF estimated an average T2 for white matter of 77ms (the ground truth was 74ms) versus 50 ms of MRF. For the same example the standard deviation was reduced from 18 ms to 6ms. The corrections achieved with the proposed SOC-MRF may expand the potential applications of bSSFP-MRF, taking advantage of its better SNR property.

  View details for DOI 10.1109/TMI.2021.3100293

  View details for PubMedID 34310296

• MAPL1: q-space reconstruction using ℓ1 -regularized mean apparent propagator. Magnetic resonance in medicine Varela-Mattatall, G., Castillo-Passi, C., Koch, A., Mura, J., Stirnberg, R., Uribe, S., Tejos, C., Stöcker, T., Irarrazaval, P. 2020; 84 (4): 2219-2230

  Abstract

  To improve the quality of mean apparent propagator (MAP) reconstruction from a limited number of q-space samples.We implement an ℓ1 -regularised MAP (MAPL1) to consider higher order basis functions and to improve the fit without increasing the number of q-space samples. We compare MAPL1 with the least-squares optimization subject to non-negativity (MAP), and the Laplacian-regularized MAP (MAPL). We use simulations of crossing fibers and compute the normalized mean squared error (NMSE) and the Pearson's correlation coefficient to evaluate the reconstruction quality in q-space. We also compare coefficient-based diffusion indices in the simulations and in in vivo data.Results indicate that MAPL1 improves NMSE in 1 to 3% when compared to MAP or MAPL in a high undersampling regime. Additionally, MAPL1 produces more reproducible and accurate results for all sampling rates when there are enough basis functions to meet the sparsity criterion for the regularizer. These improved reconstructions also produce better coefficient-based diffusion indices for in vivo data.Adding an ℓ1 regularizer to MAP allows the use of more basis functions and a better fit without increasing the number of q-space samples. The impact of our research is that a complete diffusion spectrum can be reconstructed from an acquisition time very similar to a diffusion tensor imaging protocol.

  View details for DOI 10.1002/mrm.28268

  View details for PubMedID 32270542