Filtern
Volltext vorhanden
- ja (2)
Gehört zur Bibliographie
- ja (2)
Dokumenttyp
- Artikel / Aufsatz in einer Zeitschrift (2) (entfernen)
Sprache
- Englisch (2) (entfernen)
Schlagworte
- cardiovascular magnetic resonance (CMR) (1)
- cone-beamcomputed tomography (1)
- deep learning (1)
- extremities (1)
- fractures, bone (1)
- radial (1)
- radiography (1)
- semantic segmentation (1)
- undersampling (1)
Institut
- Institut für diagnostische und interventionelle Radiologie (Institut für Röntgendiagnostik) (2) (entfernen)
Objectives
Trauma evaluation of extremities can be challenging in conventional radiography. A multi-use x-ray system with cone-beam CT (CBCT) option facilitates ancillary 3-D imaging without repositioning. We assessed the clinical value of CBCT scans by analyzing the influence of additional findings on therapy.
Methods
Ninety-two patients underwent radiography and subsequent CBCT imaging with the twin robotic scanner (76 wrist/hand/finger and 16 ankle/foot/toe trauma scans). Reports by on-call radiologists before and after CBCT were compared regarding fracture detection, joint affliction, comminuted injuries, and diagnostic confidence. An orthopedic surgeon recommended therapy based on reported findings. Surgical reports (N = 52) and clinical follow-up (N = 85) were used as reference standard.
Results
CBCT detected more fractures (83/64 of 85), joint involvements (69/53 of 71), and multi-fragment situations (68/50 of 70) than radiography (all p < 0.001). Six fractures suspected in radiographs were ruled out by CBCT. Treatment changes based on additional information from CBCT were recommended in 29 patients (31.5%). While agreement between advised therapy before CBCT and actual treatment was moderate (κ = 0.41 [95% confidence interval 0.35–0.47]; p < 0.001), agreement after CBCT was almost perfect (κ = 0.88 [0.83–0.93]; p < 0.001). Diagnostic confidence increased considerably for CBCT studies (p < 0.001). Median effective dose for CBCT was 4.3 μSv [3.3–5.3 μSv] compared to 0.2 μSv [0.1–0.2 μSv] for radiography.
Conclusions
CBCT provides advantages for the evaluation of acute small bone and joint trauma by detecting and excluding extremity fractures and fracture-related findings more reliably than radiographs. Additional findings induced therapy change in one third of patients, suggesting substantial clinical impact.
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.