RMRobin A. Müller

Robin A. Müller

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Robin A. Müller has a total of 29 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Effects of Single and Multiple Dose Tiotropium/Olodaterol in Chronic Obstructive Pulmonary Disease on Alveolar Gas Exchange Measured by Hyperpolarized 129 Xe MRI

    by Agilo L. Kern, Thomas Kayser, Robin A. Müller, Andreas Voskrebenzev, Filip Klimeš, Frank Wacker, Jens M. Hohlfeld, Jens Vogel‐Claussen

    Background: Long‐acting muscarinic antagonists/long‐acting beta adrenoreceptor agonists lead to lung deflation and improved microvascular perfusion in hyperinflated chronic obstructive pulmonary disease (COPD), but with an unclear time course. Purpose: To investigate the effect of inhalation of tiotropium/olodaterol (TIO/OLO) on MRI measures of gas uptake and microvascular blood volume after single and multiple doses. Study Type: Prospective, randomized, investigator‐blinded, placebo‐controlled single‐dose crossover study with open‐label multiple‐dose extension. Population: 31 participants with COPD and lung hyperinflation (residual volume > 135% predicted, ≥ 10 pack‐years smoking). Fieldstrength/Sequence: 1.5 T/Interleaved 3D radial gas and dissolved‐phase MRI using multi‐echo FLASH, high‐resolution spectroscopy, chemical shift saturation recovery spectroscopy. Assessment: Ratios of 129 Xe in membrane tissues (M) to gas (M‐Gas), red blood cells (RBC) to gas (RBC‐Gas), RBC to M (RBC‐M) were calculated from dissolved‐phase imaging. Dissolved‐phase apparent transverse relaxation times (T 2 *) from high‐resolution spectra were also calculated, and septal wall thickness and surface‐volume ratio (S/V) were assessed by fitting a model of xenon exchange functions. Parameters were compared for single‐dose (5 μg/5 μg TIO/OLO) versus placebo using saline, and single‐dose versus multiple‐dose (daily 5 μg/5 μg TIO/OLO for 14 days) treatments. Statistical Tests: Linear mixed models with contrasts of estimated marginal means, significance level 0.05. Results: Compared with placebo, single‐dose treatment significantly increased RBC‐Gas and septal thickness but not RBC‐M ( p = 0.06), M‐Gas ( p = 0.16), RBC T 2 * ( p = 0.13), M T 2 * ( p = 0.09) and S/V ( p = 0.47). Multiple‐dose treatment significantly increased RBC‐Gas, M‐Gas and M T 2 * compared with single‐dose treatment, but not RBC‐M ( p = 0.82), RBC T 2 * ( p = 0.48), septal thickness ( p = 0.61), and S/V ( p = 0.49). Data Conclusion: Single‐dose TIO/OLO increased 129 Xe in RBC‐Gas. Multiple‐dose treatment increased M‐Gas and further increased RBC‐Gas compared to single‐dose treatment. Technical Efficacy: Stage 4. Evidence Level: 2.


  • Open Access

    Accelerating PREFUL MRI: Comparison of Registration Methods and Impact of Time Series Reduction

    by A. Voskrebenzev, G. H. Pöhler, F. Klimeš, M. M. Klein, R. A. Müller, F. Wacker, J. M. Hohlfeld, J. Vogel‐Claussen

    The purpose of this study is to evaluate whether accelerated registration and reduced acquisition times preserve the robustness of phase‐resolved functional lung (PREFUL) MRI, thereby improving clinical feasibility by facilitating integration into tight time slots and enabling faster turnaround of results. PREFUL MRI at 1.5 T was retrospectively analyzed in 28 patients with chronic obstructive pulmonary disease (COPD) and 56 healthy volunteers. Image registration was performed using Advanced Normalization Tools (ANTs) and Forsberg. Time series (TS) were truncated from 55 to 45, 30, and 15 s per slice. The 55‐s variant was evaluated twice to assess potential nondeterministic effects; the first served as the reference. PREFUL parameters, including regional ventilation, quantified perfusion (QQ), flow volume loop correlation metric (FVL‐CM) and corresponding defect percentages, were compared across registration methods and TS lengths using Wilcoxon tests, Pearson correlations, intraclass correlation coefficients (ICC), and regional overlap analysis. Forsberg reduced computation time by ~20‐fold (3 vs. 58 min per slice) while maintaining trends and repeatability. Although functional values were slightly lower and defect scores higher, correlations between registrations exceeded 90% for most parameters, with regional overlap ≥ 97% in healthy volunteers and ≥ 80% in COPD patients. Shortening the TS to 30 s introduced minor deviations but preserved overlap (≥ 79%) and strong correlations (≥ 77%) compared with 55 s, except for QQ. Larger deviations occurred primarily at 15 s. Repeatability remained moderate to high (ICC ≥ 68%) for all parameters, except QQ, down to 30 s, with the FVL‐CM, perfusion defect percentage and amount of nondefect regions in COPD being most stable (ICC ≥ 72%). Forsberg provides results comparable to ANTs at a fraction of the processing time, enabling PREFUL analysis within a practical timeframe. Acquisition for each slice may be reduced to 30 s without compromising robustness, supporting more efficient protocols, although caution is warranted for QQ.


  • Open Access

    Dual‐Center Repeatability, Consistency, and Comparison of 3D Phase‐Resolved Functional (PREFUL) Ventilation MRI at 3 and 1.5 T in Healthy Volunteers

    by Filip Klimeš, Chuan T. Foo, Marcel Gutberlet, Andreas Voskrebenzev, Richard McIntyre, Marius M. Wernz, Norman Kornemann, Rimma Kondrashova, Robin A. Müller, Robert Grimm, et al.

    3D phase‐resolved functional lung (PREFUL) MRI offers lung ventilation assessment without contrast and in free breathing. However, there are limited data on its repeatability and intercenter consistency. This study aimed to assess the repeatability and intercenter consistency of 3D PREFUL MRI at two centers using both 3‐ and 1.5‐T field strengths. Fifty healthy volunteers (35 female, 70%; median age 25.5 [interquartile range 22.3–29.0 years]) participated in this prospective study. A golden‐angle stack‐of‐stars spoiled gradient echo acquisition was used for 3D PREFUL MRI. To assess for repeatability, 3D PREFUL MRI was performed twice with a short interval between scans at each center. At Center 1, scans were performed at 3 and 1.5 T. At Center 2, scans were performed at 3 T. Intercenter consistency of 3D PREFUL MRI was assessed between centers at 3 T. Performance of 3D PREFUL MRI at 3 and 1.5 T was compared at Center 1. Bland–Altman analysis, coefficient of variation, intraclass correlation coefficient, and Cohen's d were used for the repeatability assessment. Similarly, interfield strength and intercenter differences were quantified by Bland–Altman analysis. In all comparisons, paired Wilcoxon signed‐rank test was used to examine differences. 3D PREFUL MRI parameters showed no significant differences between repeated measurements across three scanners, except for mean regional ventilation (RVent) at Center 2 (p = 0.0046). Apart from mean RVent during the second measurement (p = 0.03) and tidal volume during both measurements (both p < 0.02), no significant differences were found between 3D PREFUL MRI parameters at 3 T across both centers. Output parameters were significantly different when images were acquired using 3 T compared to 1.5 T (all p < 0.0093). 3D PREFUL MRI ventilation‐weighted parameters showed fair to excellent repeatability and consistent intercenter results. Given the significant differences in outputs acquired using scanners of different field strengths, multicenter studies should be conducted with scanners of the same field strength.


  • Open Access

    Artificially-generated consolidations and balanced augmentation increase performance of U-net for lung parenchyma segmentation on MR images

    by Cristian Crisosto, Andreas Voskrebenzev, Marcel Gutberlet, Filip Klimeš, Till F. Kaireit, Gesa Pöhler, Tawfik Moher, Lea Behrendt, Robin Müller, Maximilian Zubke, et al.

    Purpose: To improve automated lung segmentation on 2D lung MR images using balanced augmentation and artificially-generated consolidations for training of a convolutional neural network (CNN). Materials and methods: From 233 healthy volunteers and 100 patients, 1891 coronal MR images were acquired. Of these, 1666 images without consolidations were used to build a binary semantic CNN for lung segmentation and 225 images (187 without consolidations, 38 with consolidations) were used for testing. To increase CNN performance of segmenting lung parenchyma with consolidations, balanced augmentation was performed and artificially-generated consolidations were added to all training images. The proposed CNN (CNNBal/Cons) was compared to two other CNNs: CNNUnbal/NoCons—without balanced augmentation and artificially-generated consolidations and CNNBal/NoCons—with balanced augmentation but without artificially-generated consolidations. Segmentation results were assessed using Sørensen-Dice coefficient (SDC) and Hausdorff distance coefficient. Results: Regarding the 187 MR test images without consolidations, the mean SDC of CNNUnbal/NoCons (92.1 ± 6% (mean ± standard deviation)) was significantly lower compared to CNNBal/NoCons (94.0 ± 5.3%, P = 0.0013) and CNNBal/Cons (94.3 ± 4.1%, P = 0.0001). No significant difference was found between SDC of CNNBal/Cons and CNNBal/NoCons (P = 0.54). For the 38 MR test images with consolidations, SDC of CNNUnbal/NoCons (89.0 ± 7.1%) was not significantly different compared to CNNBal/NoCons (90.2 ± 9.4%, P = 0.53). SDC of CNNBal/Cons (94.3 ± 3.7%) was significantly higher compared to CNNBal/NoCons (P = 0.0146) and CNNUnbal/NoCons (P = 0.001). Conclusions: Expanding training datasets via balanced augmentation and artificially-generated consolidations improved the accuracy of CNNBal/Cons, especially in datasets with parenchymal consolidations. This is an important step towards a robust automated postprocessing of lung MRI datasets in clinical routine.