Author name cluster

Martin Bendszus

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

9 papers

1 author row

YNIMG Journal 2024 Journal Article

Perfusion parameter map generation from TOF-MRA in stroke using generative adversarial networks

Felix Lohrke
Vince Istvan Madai
Tabea Kossen
Orhun Utku Aydin
Jonas Behland
Adam Hilbert
Matthias Anthony Mutke
Martin Bendszus

PURPOSE: To generate perfusion parameter maps from Time-of-flight magnetic resonance angiography (TOF-MRA) images using artificial intelligence to provide an alternative to traditional perfusion imaging techniques. MATERIALS AND METHODS: This retrospective study included a total of 272 patients with cerebrovascular diseases; 200 with acute stroke (from 2010 to 2018), and 72 with steno-occlusive disease (from 2011 to 2014). For each patient the TOF MRA image and the corresponding Dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) were retrieved from the datasets. The authors propose an adapted generative adversarial network (GAN) architecture, 3D pix2pix GAN, that generates common perfusion maps (CBF, CBV, MTT, TTP, Tmax) from TOF-MRA images. The performance was evaluated by the structural similarity index measure (SSIM). For a subset of 20 patients from the acute stroke dataset, the Dice coefficient was calculated to measure the overlap between the generated and real hypoperfused lesions with a time-to-maximum (Tmax) > 6 s. RESULTS: The GAN model exhibited high visual overlap and performance for all perfusion maps in both datasets: acute stroke (mean SSIM 0.88-0.92, mean PSNR 28.48-30.89, mean MAE 0.02-0.04 and mean NRMSE 0.14-0.37) and steno-occlusive disease patients (mean SSIM 0.83-0.98, mean PSNR 23.62-38.21, mean MAE 0.01-0.05 and mean NRMSE 0.03-0.15). For the overlap analysis for lesions with Tmax>6 s, the median Dice coefficient was 0.49. CONCLUSION: Our AI model can successfully generate perfusion parameter maps from TOF-MRA images, paving the way for a non-invasive alternative for assessing cerebral hemodynamics in cerebrovascular disease patients. This method could impact the stratification of patients with cerebrovascular diseases. Our results warrant more extensive refinement and validation of the method.

Details DOI

YNIMG Journal 2022 Journal Article

A novel gradient echo data based vein segmentation algorithm and its application for the detection of regional cerebral differences in venous susceptibility

Sina Straub
Janis Stiegeler
Edris El-Sanosy
Martin Bendszus
Mark E. Ladd
Till M. Schneider

Accurate segmentation of cerebral venous vasculature from gradient echo data is of central importance in several areas of neuroimaging such as for the susceptibility-based assessment of brain oxygenation or planning of electrode placement in deep brain stimulation. In this study, a vein segmentation algorithm for single- and multi-echo gradient echo data is proposed. First, susceptibility maps, true susceptibility-weighted images, and, in the multi-echo case, R 2 * maps were generated from the gradient echo data. These maps were filtered with an inverted Hamming filter to suppress background contrast as well as artifacts from field inhomogeneities at the brain boundaries. A shearlet-based scale-wise representation was generated to calculate a vesselness function and to generate segmentations based on local thresholding. The accuracy of the proposed algorithm was evaluated for different echo times and image resolutions using a manually generated reference segmentation and two vein segmentation algorithms (Frangi vesselness-based, recursive vesselness filter) as a reference with the Dice and Cohen's coefficients as well as the modified Hausdorff distance. The Frangi-based and recursive vesselness filter methods were significantly outperformed with regard to all error metrics. Applying the algorithm, susceptibility differences likely related to differences in blood oxygenation between superficial and deep venous territories could be demonstrated.

Details DOI

YNICL Journal 2020 Journal Article

Ultra-high-field sodium MRI as biomarker for tumor extent, grade and IDH mutation status in glioma patients

Sebastian Regnery
Nicolas G.R. Behl
Tanja Platt
Nina Weinfurtner
Paul Windisch
Katerina Deike-Hofmann
Felix Sahm
Martin Bendszus

PURPOSE: Na) MRI at 7 Tesla (T) field strength as biomarker for tumor extent, isocitrate dehydrogenase (IDH) mutation and O6-methylguanine DNA methyltransferase (MGMT) promotor methylation in glioma patients. METHODS: Na concentration of this whole tumor area was compared between IDH-mutated and IDH wild-type gliomas as well as MGMT methylated and MGMT not-methylated glioblastomas using Whitney-Mann U-tests. All p-values were corrected after the Bonferroni-Holm procedure. RESULTS: Na concentration in MGMT methylated tumors that did not reach statistical significance (median [interquartile range]: MGMT methylated = 57.59 mM [50.70 - 59.17 mM], MGMT not methylated = 48.78 mM [45.88 - 53.91 mM], p = 1.0). CONCLUSIONS: Na concentration from central necrosis to normal-appearing white matter suggests a correlation with tumor infiltration.

Details DOI

YNICL Journal 2019 Journal Article

Diffusion tensor imaging in anterior interosseous nerve syndrome – functional MR Neurography on a fascicular level

Tim Godel
Mirko Pham
Henrich Kele
Moritz Kronlage
Daniel Schwarz
Merle Brunée
Sabine Heiland
Martin Bendszus

PURPOSE: By applying diffusor tensor imaging (DTI) in patients with anterior interosseous nerve syndrome (AINS), this proof of principle study aims to quantify the extent of structural damage of a peripheral nerve at the anatomical level of individual fascicles. METHODS: In this institutional review board approved prospective study 13 patients with spontaneous AINS were examined at 3 Tesla including a transversal T2-weighted turbo-spin-echo and a spin-echo echo-planar-imaging pulse sequence of the upper arm level. Calculations of quantitative DTI parameters including fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD) for median nerve lesion and non-lesion fascicles as well as ulnar and radial nerve were obtained. DTI values were compared to each other and to a previously published dataset of 58 healthy controls using one-way Analysis of Variance with Bonferroni correction and p-values <.05 were considered significant. Receiver operating characteristic (ROC) curves were performed to assess diagnostic accuracy. RESULTS: FA of median nerve lesion fascicles was decreased compared to median nerve non-lesion fascicles, ulnar nerve and radial nerve while MD, RD, and AD was increased (p < .001 for all parameters). Compared to median nerve values of healthy controls, lesion fascicles showed a significant decrease in FA while MD, RD, and AD was increased (p < .001 for all parameters). FA of median nerve non-lesion fascicles showed a weak significant decrease compared to healthy controls (p < .01) while there was no difference in MD, RD, and AD. ROC analyses revealed an excellent diagnostic accuracy of FA, MD and RD in the discrimination of median nerve lesion and non-lesion fascicles in AINS patients as well as in the discrimination of lesion fascicles and normative median nerve values of healthy controls. CONCLUSION: By applying this functional MR Neurography technique in patients with AINS, this proof of principle study demonstrates that diffusion tensor imaging is feasible to quantify structural nerve injury at the anatomical level of individual fascicles.

Details DOI

YNIMG Journal 2016 Journal Article

Deep MRI brain extraction: A 3D convolutional neural network for skull stripping

Jens Kleesiek
Gregor Urban
Alexander Hubert
Daniel Schwarz
Klaus Maier-Hein
Martin Bendszus
Armin Biller

Brain extraction from magnetic resonance imaging (MRI) is crucial for many neuroimaging workflows. Current methods demonstrate good results on non-enhanced T1-weighted images, but struggle when confronted with other modalities and pathologically altered tissue. In this paper we present a 3D convolutional deep learning architecture to address these shortcomings. In contrast to existing methods, we are not limited to non-enhanced T1w images. When trained appropriately, our approach handles an arbitrary number of modalities including contrast-enhanced scans. Its applicability to MRI data, comprising four channels: non-enhanced and contrast-enhanced T1w, T2w and FLAIR contrasts, is demonstrated on a challenging clinical data set containing brain tumors (N=53), where our approach significantly outperforms six commonly used tools with a mean Dice score of 95. 19. Further, the proposed method at least matches state-of-the-art performance as demonstrated on three publicly available data sets: IBSR, LPBA40 and OASIS, totaling N=135 volumes. For the IBSR (96. 32) and LPBA40 (96. 96) data set the convolutional neuronal network (CNN) obtains the highest average Dice scores, albeit not being significantly different from the second best performing method. For the OASIS data the second best Dice (95. 02) results are achieved, with no statistical difference in comparison to the best performing tool. For all data sets the highest average specificity measures are evaluated, whereas the sensitivity displays about average results. Adjusting the cut-off threshold for generating the binary masks from the CNN's probability output can be used to increase the sensitivity of the method. Of course, this comes at the cost of a decreased specificity and has to be decided application specific. Using an optimized GPU implementation predictions can be achieved in less than one minute. The proposed method may prove useful for large-scale studies and clinical trials.

Details DOI

YNIMG Journal 2016 Journal Article

Human dorsal-root-ganglion perfusion measured in-vivo by MRI

Tim Godel
Mirko Pham
Sabine Heiland
Martin Bendszus
Philipp Bäumer

Purpose To develop an in-vivo imaging method for the measurement of dorsal-root-ganglia-(DRG) perfusion, to establish its normal values in patients without known peripheral nerve disorders or radicular pain syndromes and to determine the physiological spatial perfusion pattern within the DRG. Methods This prospective study was approved by the institutional ethics committee and written informed consent was obtained from all participants. 46 (24 female, 22 male, mean age 46. 0±15. 2years) subjects without known peripheral neuropathies or pain syndromes were examined by a 3Tesla MRI scanner (Magnetom VERIO or TRIO, Siemens AG, Erlangen, Germany) with a VIBE (Volume-Interpolated-Breathhold-Examination) dynamic-contrast-enhanced (DCE) T1-w-sequence (TR/TE 3. 3/1. 11ms; 24 slices; voxel resolution 1. 3×1. 3×3. 0mm3) covered the pelvis from the upper plate of the 5th lumbar vertebra to the 2nd sacral vertebra. Transfer-constant (Ktrans) and interstitial-volume-fraction (interstitial-leakage-fraction, Ve) were modeled for the DRG and spinal nerve by applying the Tofts-model. Statistical analyses included pairwise comparisons of L5/S1 DRG vs. spinal nerve. Furthermore, distinct physiological zones within the S1 DRG were compared (cell body rich area (CBRA) vs. nerve fiber rich area (NFRA)). Results DRG showed a significantly increased permeability compared to spinal nerve (Ktrans 3. 8±1. 5 10−3/min vs. 1. 6±0. 9 10−3/min, p-value: <0. 001) combined with an increased interstitial leakage of contrast agent into the extravascular-extracellular-space (Ve 38. 1±19. 2% vs. 17. 3±9. 9%, p-value: <0. 001). Parameters showed no statistically significant difference on DRG-level (L5 vs. S1; p-value: 0. 62 (Ktrans); 0. 17 (Ve)) and -side (left vs. right; p-value: 0. 25 (Ktrans); 0. 79 (Ve)). Female gender was associated with a significantly increased permeability (Ktrans female 4. 3±1. 4 10−3/min vs. male 3. 4±0. 9 10−3/min, p-value: <0. 05) but no statistically significant differences in interstitial leakage (Ve female 40. 1±14, 1% vs. male 34. 5±17. 4%, p-value: 0. 24). DRG showed distinct spatial distribution patterns of perfusion: Ktrans and Ve were significantly higher in the CBRA than in the NFRA (Ktrans 4. 4±1. 8 10−3/min vs. 1. 7±1. 2 10−3/min, p-value: <0. 001 and Ve 40. 9±21. 3% vs. 15. 1±11. 7%, p-value: <0. 001). Conclusion Non-invasive and in-vivo measurement of human DRG perfusion by MRI is a feasible technique. DRG show substantially higher permeability and interstitial leakage than spinal nerves. Even distinct physiological perfusion patterns for different microstructural compartments could be observed within the DRG. The technique may become particularly useful for future research on the poorly understood human sensory neuropathies and pain syndromes.

Details DOI

YNIMG Journal 2015 Journal Article

Relaxation-compensated CEST-MRI of the human brain at 7 T: Unbiased insight into NOE and amide signal changes in human glioblastoma

Moritz Zaiss
Johannes Windschuh
Daniel Paech
Jan-Eric Meissner
Sina Burth
Benjamin Schmitt
Philip Kickingereder
Benedikt Wiestler

Endogenous chemical exchange saturation transfer (CEST) effects of protons resonating near to water protons are always diluted by competing effects such as direct water saturation and semi-solid magnetization transfer (MT). This leads to unwanted T2 and MT signal contributions that contaminate the observed CEST signal. Furthermore, all CEST effects appear to be scaled by the T1 relaxation time of the mediating water pool. As MT, T1 and T2 are also altered in tumor regions, a recently published correction algorithm yielding the apparent exchange-dependent relaxation AREX, is used to evaluate in vivo CEST effects. This study focuses on CEST effects of amides (3. 5ppm) and Nuclear-Overhauser-mediated saturation transfer (NOE, −3. 5ppm) that can be properly isolated at 7T. These were obtained in 10 glioblastoma patients, and this is the first comprehensive study where AREX is applied in human brain as well as in human glioblastoma. The correction of CEST effects alters the contrast significantly: after correction, the CEST effect of amides does not show significant contrast between contrast enhancing tumor regions and normal tissue, whereas NOE drops significantly in the tumor area. In addition, new features in the AREX contrasts are visible. This suggests that previous CEST approaches might not have shown pure CEST effects, but rather water relaxation shine-through effects. Our insights help to improve understanding of the CEST effect changes in tumors and correlations on a cellular and molecular level.

Details DOI

YNIMG Journal 2012 Journal Article

Grey matter abnormalities within cortico-limbic-striatal circuits in acute and weight-restored anorexia nervosa patients

Hans-Christoph Friederich
Stephan Walther
Martin Bendszus
Armin Biller
Philipp Thomann
Susanne Zeigermann
Tobias Katus
Romuald Brunner

Functional disturbances within cortico-striatal control systems have been implicated in the psychobiology (i. e. impaired cognitive–behavioral flexibility, perfectionist personality) of anorexia nervosa. The aim of the present study was to investigate the morphometry of brain regions within cortico-striatal networks in acute anorexia nervosa (AN) as well as long-term weight-restored anorexia nervosa (AN-WR) patients. A total of 39 participants: 12 AN, 13 AN-WR patients, and 14 healthy controls (HC) underwent high-resolution, T1-weighted magnetic resonance imaging (MRI), a cognitive–behavioral flexibility task, and a psychometric assessment. Group differences in local grey matter volume (GMV) were analyzed using whole brain voxel-based morphometry (VBM) and brain-atlas based automatic volumetry computation (IBASPM). Individual differences in total GMV were considered as a covariate in all analyses. In the regional brain morphometry, AN patients, as compared to HC, showed decreased GMVs (VBM and volumetry) in the anterior cingulate cortex (ACC), the supplementary motor area (SMA), and in subcortical regions (amygdala, putamen: VBM only). AN-WR compared to HC showed decreased GMV (VBM and volumetry) in the ACC and SMA, whereas GMV of the subcortical region showed no differences. The findings of the study suggest that structural abnormalities of the ACC and SMA were independent of the disease stage, whereas subcortical limbic–striatal changes were state dependent.

Details DOI

YNIMG Journal 2007 Journal Article

Scanning for the scanner: FMRI of audition by read-out omissions from echo-planar imaging

Andreas J. Bartsch
György Homola
Stefan Thesen
Peter Sahmer
Ralph Keim
Christian F. Beckmann
Armin Biller
Christoph Knaus

Echo-planar imaging (EPI) generates considerable acoustic noise by rapidly oscillating gradients. In functional magnetic resonance imaging (FMRI), unshielded EPI sounds activate the auditory system inasmuch as it is responsive. Instead of attenuating EPI noise, our goal was to utilize it for auditory FMRI by omitting read-outs from the pulse sequence's gradient train. Read-out gradient pulses are the primary noise determinant of EPI introducing its peak sound level and fundamental frequency peak which inversely relates to twice the echo spacing. Using model-driven analyses, we demonstrate that withholding read-outs from EPI is suited to reliably evoke hemodynamic blood oxygenation level-dependent (BOLD) signal modulations bilaterally in the auditory cortex of normal hearing subjects (n =60). To investigate the utility of EPI read-out omissions for auditory FMRI at an individual subject's level, we compare traditional Family-Wise-Error-Rate (FWER)-corrected maximum height thresholding to spatial mixture modeling (SMM). With the latter, appropriate bilateral auditory activations were confirmed in 95% of the individuals, whereas FWER-based voxel thresholding detected such activations in up to 72%. We illustrate the applicability of this novel EPI modification for clinical diagnostic purposes and report on a patient with bilateral large vestibular aqueducts (LVAs) and severe binaural sensorineural hearing loss (SNHL). In this particular case, read-out omissions from EPI were used to assert residual audition prior to cochlear implantation (CI). Requiring no specific task compliance or sophisticated stimulation equipment other than the scanner on its own, FMRI by read-out omissions lends itself to auditory investigations and to quickly probe audition.

Details DOI