Refine
Is part of the Bibliography
- yes (121)
Year of publication
Document Type
- Journal article (113)
- Doctoral Thesis (7)
- Preprint (1)
Language
- English (121) (remove)
Keywords
- magnetic resonance imaging (10)
- MRI (8)
- deep learning (4)
- Bildgebendes Verfahren (3)
- Medizin (3)
- cardiac magnetic resonance imaging (3)
- endovascular (3)
- lung (3)
- multiple myeloma (3)
- rotator cuff (3)
Institute
- Institut für diagnostische und interventionelle Radiologie (Institut für Röntgendiagnostik) (121) (remove)
Sonstige beteiligte Institutionen
- Cheng Lab, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA (1)
- Datenintegrationszentrum Würzburg (DIZ) (1)
- Interdisziplinäre Biomaterial- und Datenbank Würzburg (ibdw) (1)
- Würzburg Fabry Center for Interdisciplinary Therapy (FAZIT), University of Würzburg, Würzburg, Germany (1)
Acceleration is a central aim of clinical and technical research in magnetic resonance imaging (MRI) today, with the potential to increase robustness, accessibility and patient comfort, reduce cost, and enable entirely new kinds of examinations. A key component in this endeavor is image reconstruction, as most modern approaches build on advanced signal and image processing. Here, deep learning (DL)-based methods have recently shown considerable potential, with numerous publications demonstrating benefits for MRI reconstruction. However, these methods often come at the cost of an increased risk for subtle yet critical errors. Therefore, the aim of this thesis is to advance DL-based MRI reconstruction, while ensuring high quality and fidelity with measured data. A network architecture specifically suited for this purpose is the variational network (VN). To investigate the benefits these can bring to non-Cartesian cardiac imaging, the first part presents an application of VNs, which were specifically adapted to the reconstruction of accelerated spiral acquisitions. The proposed method is compared to a segmented exam, a U-Net and a compressed sensing (CS) model using qualitative and quantitative measures. While the U-Net performed poorly, the VN as well as the CS reconstruction showed good output quality. In functional cardiac imaging, the proposed real-time method with VN reconstruction substantially accelerates examinations over the gold-standard, from over 10 to just 1 minute. Clinical parameters agreed on average.
Generally in MRI reconstruction, the assessment of image quality is complex, in particular for modern non-linear methods. Therefore, advanced techniques for precise evaluation of quality were subsequently demonstrated.
With two distinct methods, resolution and amplification or suppression of noise are quantified locally in each pixel of a reconstruction. Using these, local maps of resolution and noise in parallel imaging (GRAPPA), CS, U-Net and VN reconstructions were determined for MR images of the brain. In the tested images, GRAPPA delivers uniform and ideal resolution, but amplifies noise noticeably. The other methods adapt their behavior to image structure, where different levels of local blurring were observed at edges compared to homogeneous areas, and noise was suppressed except at edges. Overall, VNs were found to combine a number of advantageous properties, including a good trade-off between resolution and noise, fast reconstruction times, and high overall image quality and fidelity of the produced output. Therefore, this network architecture seems highly promising for MRI reconstruction.
This work deals with the acceleration of cardiovascular MRI for the assessment
of functional information in steady-state contrast and for viability assessment
during the inversion recovery of the magnetization. Two approaches
are introduced and discussed in detail. MOCO-MAP uses an exponential
model to recover dynamic image data, IR-CRISPI, with its low-rank plus
sparse reconstruction, is related to compressed sensing.
MOCO-MAP is a successor to model-based acceleration of parametermapping
(MAP) for the application in the myocardial region. To this end, it
was augmented with a motion correction (MOCO) step to allow exponential
fitting the signal of a still object in temporal direction. Iteratively, this
introduction of prior physical knowledge together with the enforcement of
consistency with the measured data can be used to reconstruct an image
series from distinctly shorter sampling time than the standard exam (< 3 s
opposed to about 10 s). Results show feasibility of the method as well as
detectability of delayed enhancement in the myocardium, but also significant
discrepancies when imaging cardiac function and artifacts caused already by
minor inaccuracy of the motion correction.
IR-CRISPI was developed from CRISPI, which is a real-time protocol
specifically designed for functional evaluation of image data in steady-state
contrast. With a reconstruction based on the separate calculation of low-rank
and sparse part, it employs a softer constraint than the strict exponential
model, which was possible due to sufficient temporal sampling density via
spiral acquisition. The low-rank plus sparse reconstruction is fit for the use on
dynamic and on inversion recovery data. Thus, motion correction is rendered
unnecessary with it.
IR-CRISPI was equipped with noise suppression via spatial wavelet filtering.
A study comprising 10 patients with cardiac disease show medical
applicability. A comparison with performed traditional reference exams offer
insight into diagnostic benefits. Especially regarding patients with difficulty
to hold their breath, the real-time manner of the IR-CRISPI acquisition provides
a valuable alternative and an increase in robustness.
In conclusion, especially with IR-CRISPI in free breathing, a major acceleration
of the cardiovascular MR exam could be realized. In an acquisition
of less than 100 s, it not only includes the information of two traditional
protocols (cine and LGE), which take up more than 9.6 min, but also allows
adjustment of TI in retrospect and yields lower artifact level with similar
image quality.
Pulmonary artery embolism (PE) is a common condition and an even more common clinical suspect. The computed tomography pulmonary angiogram (CTPA) is the main medical imaging tool used to diagnose a suspected case of PE. To gain a better impression of the effects of a PE on the perfusion and hence the gas exchange, a functional imaging method is beneficial. One approach for functional imaging using radiation exposure is the generation of color-coded iodine perfusion maps acquired by Dual-Energy Computed Tomography (DECT), which enable the detection of perfusion defects in the pulmonary parenchyma. In contrast to the existing approach of DECT with iodine color-coded maps, the SElf-gated Non-Contrast-Enhanced FUnctional Lung (SENCEFUL) MRI technique offers the possibility to interpret perfusion maps without any radiation exposure or application of contrast agents. The measurement in SENCEFUL MRI can be performed during conditions of free breathing and without electrocardiogram triggering.
The purpose of this study was to determine whether PE can be diagnosed on the basis of visible perfusion defects in the perfusion maps of SENCEFUL MRI and in the iodine-coded maps of DECT and to compare the diagnostic performance of these methods. Both SENCEFUL-MRI and iodine distribution maps from DECT have been compared with the CTPA of ten patients with PE. Additionally, the functional images were compared with each other on a per-patient basis.
The iodine perfusion maps of DECT had a sensitivity of 84.2 % and specificity of 65.2 % for the diagnosis of PE. The SENCEFUL technique in MRI showed a sensitivity of 78.9 % and a specificity of 26.1 %. When comparing the whole lung depicted in both series of functional images, the main perfusion defect location matched in four of ten patients (40 %).
In conclusion, this work found that DECT iodine maps have higher sensitivity and specificity in the diagnosis of pulmonary embolism compared with SENCEFUL MRI.
Purpose
The reliable detection of tumor-infiltrated axillary lymph nodes for breast cancer [BC] patients plays a decisive role in further therapy. We aimed to find out whether cross-sectional imaging techniques could improve sensitivity for pretherapeutic axillary staging in nodal-positive BC patients compared to conventional imaging such as mammography and sonography.
Methods
Data for breast cancer patients with tumor-infiltrated axillary lymph nodes having received surgery between 2014 and 2020 were included in this study.
All examinations (sonography, mammography, computed tomography [CT] and magnetic resonance imaging [MRI]) were interpreted by board-certified specialists in radiology. The sensitivity of different imaging modalities was calculated, and binary logistic regression analyses were performed to detect variables influencing the detection of positive lymph nodes.
Results
All included 382 breast cancer patients had received conventional imaging, while 52.61% of the patients had received cross-sectional imaging.
The sensitivity of the combination of all imaging modalities was 68.89%. The combination of MRI and CT showed 63.83% and the combination of sonography and mammography showed 36.11% sensitivity.
Conclusion
We could demonstrate that cross-sectional imaging can improve the sensitivity of the detection of tumor-infiltrated axillary lymph nodes in breast cancer patients. Only the safe detection of these lymph nodes at the time of diagnosis enables the evaluation of the response to neoadjuvant therapy, thereby allowing access to prognosis and improving new post-neoadjuvant therapies.
Objectives: Positron emission tomography (PET) is currently considered the non-invasive reference standard for lymph node (N-)staging in lung cancer. However, not all patients can undergo this diagnostic procedure due to high costs, limited availability, and additional radiation exposure. The purpose of this study was to predict the PET result from traditional contrast-enhanced computed tomography (CT) and to test different feature extraction strategies. Methods: In this study, 100 lung cancer patients underwent a contrast-enhanced \(^{18}\)F-fluorodeoxyglucose (FDG) PET/CT scan between August 2012 and December 2019. We trained machine learning models to predict FDG uptake in the subsequent PET scan. Model inputs were composed of (i) traditional “hand-crafted” radiomics features from the segmented lymph nodes, (ii) deep features derived from a pretrained EfficientNet-CNN, and (iii) a hybrid approach combining (i) and (ii). Results: In total, 2734 lymph nodes [555 (20.3%) PET-positive] from 100 patients [49% female; mean age 65, SD: 14] with lung cancer (60% adenocarcinoma, 21% plate epithelial carcinoma, 8% small-cell lung cancer) were included in this study. The area under the receiver operating characteristic curve (AUC) ranged from 0.79 to 0.87, and the scaled Brier score (SBS) ranged from 16 to 36%. The random forest model (iii) yielded the best results [AUC 0.871 (0.865–0.878), SBS 35.8 (34.2–37.2)] and had significantly higher model performance than both approaches alone (AUC: p < 0.001, z = 8.8 and z = 22.4; SBS: p < 0.001, z = 11.4 and z = 26.6, against (i) and (ii), respectively). Conclusion: Both traditional radiomics features and transfer-learning deep radiomics features provide relevant and complementary information for non-invasive N-staging in lung cancer.
Objectives: Dual-source dual-energy CT (DECT) facilitates reconstruction of virtual non-contrast images from contrast-enhanced scans within a limited field of view. This study evaluates the replacement of true non-contrast acquisition with virtual non-contrast reconstructions and investigates the limitations of dual-source DECT in obese patients. Materials and Methods: A total of 253 oncologic patients (153 women; age 64.5 ± 16.2 years; BMI 26.6 ± 5.1 kg/m\(^2\)) received both multi-phase single-energy CT (SECT) and DECT in sequential staging examinations with a third-generation dual-source scanner. Patients were allocated to one of three BMI clusters: non-obese: <25 kg/m\(^2\) (n = 110), pre-obese: 25–29.9 kg/m\(^2\) (n = 73), and obese: >30 kg/m\(^2\) (n = 70). Radiation dose and image quality were compared for each scan. DECT examinations were evaluated regarding liver coverage within the dual-energy field of view. Results: While arterial contrast phases in DECT were associated with a higher CTDI\(_{vol}\) than in SECT (11.1 vs. 8.1 mGy; p < 0.001), replacement of true with virtual non-contrast imaging resulted in a considerably lower overall dose-length product (312.6 vs. 475.3 mGy·cm; p < 0.001). The proportion of DLP variance predictable from patient BMI was substantial in DECT (R\(^2\) = 0.738) and SECT (R\(^2\) = 0.620); however, DLP of SECT showed a stronger increase in obese patients (p < 0.001). Incomplete coverage of the liver within the dual-energy field of view was most common in the obese subgroup (17.1%) compared with non-obese (0%) and pre-obese patients (4.1%). Conclusion: DECT facilitates a 30.8% dose reduction over SECT in abdominal oncologic staging examinations. Employing dual-source scanner architecture, the risk for incomplete liver coverage increases in obese patients.
In this study, the impact of reconstruction sharpness on the visualization of the appendicular skeleton in ultrahigh-resolution (UHR) photon-counting detector (PCD) CT was investigated. Sixteen cadaveric extremities (eight fractured) were examined with a standardized 120 kVp scan protocol (CTDI\(_{vol}\) 10 mGy). Images were reconstructed with the sharpest non-UHR kernel (Br76) and all available UHR kernels (Br80 to Br96). Seven radiologists evaluated image quality and fracture assessability. Interrater agreement was assessed with the intraclass correlation coefficient. For quantitative comparisons, signal-to-noise-ratios (SNRs) were calculated. Subjective image quality was best for Br84 (median 1, interquartile range 1–3; p ≤ 0.003). Regarding fracture assessability, no significant difference was ascertained between Br76, Br80 and Br84 (p > 0.999), with inferior ratings for all sharper kernels (p < 0.001). Interrater agreement for image quality (0.795, 0.732–0.848; p < 0.001) and fracture assessability (0.880; 0.842–0.911; p < 0.001) was good. SNR was highest for Br76 (3.4, 3.0–3.9) with no significant difference to Br80 and Br84 (p > 0.999). Br76 and Br80 produced higher SNRs than all kernels sharper than Br84 (p ≤ 0.026). In conclusion, PCD-CT reconstructions with a moderate UHR kernel offer superior image quality for visualizing the appendicular skeleton. Fracture assessability benefits from sharp non-UHR and moderate UHR kernels, while ultra-sharp reconstructions incur augmented image noise.
Introduction
In spinal surgery, precise instrumentation is essential. This study aims to evaluate the accuracy of navigated, O-arm-controlled screw positioning in thoracic and lumbar spine instabilities.
Materials and methods
Posterior instrumentation procedures between 2010 and 2015 were retrospectively analyzed. Pedicle screws were placed using 3D rotational fluoroscopy and neuronavigation. Accuracy of screw placement was assessed using a 6-grade scoring system. In addition, screw length was analyzed in relation to the vertebral body diameter. Intra- and postoperative revision rates were recorded.
Results
Thoracic and lumbar spine surgery was performed in 285 patients. Of 1704 pedicle screws, 1621 (95.1%) showed excellent positioning in 3D rotational fluoroscopy imaging. The lateral rim of either pedicle or vertebral body was protruded in 25 (1.5%) and 28 screws (1.6%), while the midline of the vertebral body was crossed in 8 screws (0.5%). Furthermore, 11 screws each (0.6%) fulfilled the criteria of full lateral and medial displacement. The median relative screw length was 92.6%. Intraoperative revision resulted in excellent positioning in 58 of 71 screws. Follow-up surgery due to missed primary malposition had to be performed for two screws in the same patient. Postsurgical symptom relief was reported in 82.1% of patients, whereas neurological deterioration occurred in 8.9% of cases with neurological follow-up.
Conclusions
Combination of neuronavigation and 3D rotational fluoroscopy control ensures excellent accuracy in pedicle screw positioning. As misplaced screws can be detected reliably and revised intraoperatively, repeated surgery for screw malposition is rarely required.
Objective
Blindness is a feared complication of giant cell arteritis (GCA). However, the spectrum of pathologic orbital imaging findings on magnetic resonance imaging (MRI) in GCA is not well understood. In this study, we assess inflammatory changes of intraorbital structures on black blood MRI (BB-MRI) in patients with GCA compared to age-matched controls.
Methods
In this multicenter case-control study, 106 subjects underwent BB-MRI. Fifty-six patients with clinically or histologically diagnosed GCA and 50 age-matched controls without clinical or laboratory evidence of vasculitis were included. All individuals were imaged on a 3-T MR scanner with a post-contrast compressed-sensing (CS) T1-weighted sampling perfection with application-optimized contrasts using different flip angle evolution (SPACE) BB-MRI sequence. Imaging results were correlated with available clinical symptoms.
Results
Eighteen of 56 GCA patients (32%) showed inflammatory changes of at least one of the intraorbital structures. The most common finding was enhancement of at least one of the optic nerve sheaths (N = 13, 72%). Vessel wall enhancement of the ophthalmic artery was unilateral in 8 and bilateral in 3 patients. Enhancement of the optic nerve was observed in one patient. There was no significant correlation between imaging features of inflammation and clinically reported orbital symptoms (p = 0.10). None of the age-matched control patients showed any inflammatory changes of intraorbital structures.
Conclusions
BB-MRI revealed inflammatory findings in the orbits in up to 32% of patients with GCA. Optic nerve sheath enhancement was the most common intraorbital inflammatory change on BB-MRI. MRI findings were independent of clinically reported orbital symptoms.
Key Points
• Up to 32% of GCA patients shows signs of inflammation of intraorbital structures on BB-MRI.
• Enhancement of the optic nerve sheath is the most common intraorbital finding in GCA patients on BB-MRI.
• Features of inflammation of intraorbital structures are independent of clinically reported symptoms.
Purpose
Image acquisition and subsequent manual analysis of cardiac cine MRI is time-consuming. The purpose of this study was to train and evaluate a 3D artificial neural network for semantic segmentation of radially undersampled cardiac MRI to accelerate both scan time and postprocessing.
Methods
A database of Cartesian short-axis MR images of the heart (148,500 images, 484 examinations) was assembled from an openly accessible database and radial undersampling was simulated. A 3D U-Net architecture was pretrained for segmentation of undersampled spatiotemporal cine MRI. Transfer learning was then performed using samples from a second database, comprising 108 non-Cartesian radial cine series of the midventricular myocardium to optimize the performance for authentic data. The performance was evaluated for different levels of undersampling by the Dice similarity coefficient (DSC) with respect to reference labels, as well as by deriving ventricular volumes and myocardial masses.
Results
Without transfer learning, the pretrained model performed moderately on true radial data [maximum number of projections tested, P = 196; DSC = 0.87 (left ventricle), DSC = 0.76 (myocardium), and DSC =0.64 (right ventricle)]. After transfer learning with authentic data, the predictions achieved human level even for high undersampling rates (P = 33, DSC = 0.95, 0.87, and 0.93) without significant difference compared with segmentations derived from fully sampled data.
Conclusion
A 3D U-Net architecture can be used for semantic segmentation of radially undersampled cine acquisitions, achieving a performance comparable with human experts in fully sampled data. This approach can jointly accelerate time-consuming cine image acquisition and cumbersome manual image analysis.