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.
Rationale and Objectives Pulmonary magnetic resonance angiography (MRA) is an imaging method with proven utility for the exclusion of pulmonary embolism and avoids the need for ionizing radiation and iodinated contrast agents. High-relaxivity gadolinium-based contrast agents (GBCAs), such as gadopiclenol, can be used to reduce the required gadolinium dose for pulmonary MRA. The aim of this study was to compare the contrast enhancement performance of gadopiclenol with an established gadobenate dimeglumine–enhanced pulmonary MRA protocol. Materials and Methods In this retrospective single-center study, data from 152 patients who underwent pulmonary MRA at 1.5 T were analyzed. Imaging was performed with either 0.05 mmol/kg gadopiclenol (n = 75) or 0.1 mmol/kg gadobenate dimeglumine (n = 77), using dedicated multiphasic imaging protocols with precontrast, pulmonary arterial phase, immediate delayed phase, and a low flip-angle T1-weighted spoiled gradient echo acquisition. Subjective image quality evaluation was performed blinded by 2 radiologists on a 5-point Likert scale. For the estimation of interrater reliability, Cohen weighted κ was calculated. For semiquantitative assessment, signal intensities were measured in the pulmonary arteries, and relative signal enhancement was calculated. Data from groups were compared with Mann-Whitney U tests using Bonferroni corrections. Results Signal enhancement relative to precontrast in the first-pass pulmonary arterial phase was higher with 0.05 mmol/kg gadopiclenol compared with 0.1 mmol/kg gadobenate dimeglumine (20.0-fold ± 5.6-fold vs 17.8-fold ± 5.8-fold; P = 0.015). Readers observed no difference in subjective rating in terms of intravascular contrast, peripheral vessel depiction, and diagnostic confidence with substantial interrater reliability (Cohen κ = 0.73 [95% confidence interval: 0.57–0.89], 0.65 [0.55–0.75], and 0.74 [0.65–0.84], all P 's < 0.001). No severe adverse events were recorded for any clinical MRA examination. Conclusions The high-relaxivity contrast agent gadopiclenol can facilitate a reduction in gadolinium dose by 50% without compromising contrast enhancement for pulmonary MRA. This approach may enhance the safety and sustainability of pulmonary MRA in the long term.
The purpose of the current study was to implement and validate joint real-time acquisition of functional and late gadolinium-enhancement (LGE) cardiac magnetic resonance (MR) images during free breathing. Inversion recovery cardiac real-time images with a temporal resolution of 50 ms were acquired using a spiral trajectory (IR-CRISPI) with a pre-emphasis based on the gradient system transfer function during free breathing. Functional and LGE cardiac MR images were reconstructed using a low-rank plus sparse model. Late gadolinium-enhancement appearance, image quality, and functional parameters of IR-CRISPI were compared with clinical standard balanced steady-state free precession breath-hold techniques in 10 patients. The acquisition of IR-CRISPI in free breathing of the entire left ventricle took 97 s on average. Bland-Altman analysis and Wilcoxon tests showed a higher artifact level for the breath-hold technique (p = 0.003), especially for arrhythmic patients or patients with dyspnea, but an increased noise level for IR-CRISPI of the LGE images (p = 0.01). The estimated transmural extent of the enhancement differed by not more than 25% and did not show a significant bias between the techniques (p = 0.50). The ascertained functional parameters were similar for the breath-hold technique and IR-CRISPI, that is, with a minor, nonsignificant (p = 0.16) mean difference of the ejection fraction of 2.3% and a 95% confidence interval from -4.8% to 9.4%. IR-CRISPI enables joint functional and LGE imaging in free breathing with good image quality but distinctly shorter scan times in comparison with breath-hold techniques.
Purpose: Ultrashort echotime (UTE) sequences aim to improve the signal yield in pulmonary magnetic resonance imaging (MRI). We demonstrate the initial results of spiral 3-dimensional (3D) UTE-MRI for combined morphologic and functional imaging in pediatric patients. Methods: Seven pediatric patients with pulmonary abnormalities were included in this observational, prospective, single-center study, with the patients having the following conditions: cystic fibrosis (CF) with middle lobe atelectasis, CF with allergic bronchopulmonary aspergillosis, primary ciliary dyskinesia, air trapping, congenital lobar overinflation, congenital pulmonary airway malformation, and pulmonary hamartoma. Patients were scanned during breath-hold in 5 breathing states on a 3-Tesla system using a prototypical 3D stack-of-spirals UTE sequence. Ventilation maps and signal intensity maps were calculated. Morphologic images, ventilation-weighted maps, and signal intensity maps of the lungs of each patient were assessed intraindividually and compared with reference examinations. Results: With a scan time of ∼15 seconds per breathing state, 3D UTE-MRI allowed for sufficient imaging of both “plus” pathologies (atelectasis, inflammatory consolidation, and pulmonary hamartoma) and “minus” pathologies (congenital lobar overinflation, congenital pulmonary airway malformation, and air trapping). Color-coded maps of normalized signal intensity and ventilation increased diagnostic confidence, particularly with regard to “minus” pathologies. UTE-MRI detected new atelectasis in an asymptomatic CF patient, allowing for rapid and successful therapy initiation, and it was able to reproduce atelectasis and hamartoma known from multidetector computed tomography and to monitor a patient with allergic bronchopulmonary aspergillosis. Conclusion: 3D UTE-MRI using a stack-of-spirals trajectory enables combined morphologic and functional imaging of the lungs within ~115 second acquisition time and might be suitable for monitoring a wide spectrum of pulmonary diseases.
• The deep learning algorithm detected intracranial hemorrhage (ICH) with 91.0% accuracy (sensitivity 91.4% specificity 90.4%). • In comparison with the assigned radiologist, the deep learning algorithm was able to highlight presence of ICH significantly faster. • In ensemble, the algorithm and the radiologist detected presence of ICH with sensitivity of 100% in a small clinical cohort. We evaluate the performance of a deep learning-based pipeline using a Dense U-net architecture for detection of intracranial hemorrhage (ICH) in unenhanced head computed tomography (CT) scans. A balanced database was assembled retrospectively, comprising a total of 872 CT scans (362 with present ICH). Predictions by the algorithm were analyzed and compared to the radiology report (ground truth). Secondly, the algorithm's performance was tested in clinical environment: A total of 100 head CT scans (11 with present ICH) were analyzed simultaneously by the deep learning algorithm and a radiologist during clinical routine. The time until first temporary diagnosis of ICH was measured. Performances of the algorithm were evaluated in combination with the radiologist, when using it as triage tool. In the retrospectively assembled dataset the deep learning algorithm detected ICH with a sensitivity of 91.4%, specificity of 90.4% and overall accuracy of 91.0%. In clinical environment, the algorithm was significantly faster compared to the temporary report of the assigned radiologist (24 ± 2 s vs. 613 ± 658 s, p < 0.001). When using the algorithm as a triage tool additional to the report of the assigned radiologist, a sensitivity of 100% was achieved. These results and the short processing time demonstrate the immense potential of deep learning applications for the use as triage tool and for additional review of manual reports.
Objectives Hardening the x-ray beam, tin prefiltration is established for imaging of high-contrast subjects in energy-integrating detector computed tomography (EID-CT). With this work, we aimed to investigate the dose-saving potential of spectral shaping via tin prefiltration in photon-counting detector CT (PCD-CT) of the temporal bone. Methods Deploying dose-matched scan protocols with and without tin prefiltration on a PCD-CT and EID-CT system (low-/intermediate-/full-dose: 4.8/7.6–7.7/27.0–27.1 mGy), 12 ultra-high-resolution examinations were performed on each of 5 cadaveric heads. While 120 kVp was applied for standard imaging, the protocols with spectral shaping used the highest potential available with tin prefiltration (EID-CT: Sn 150 kVp, PCD-CT: Sn 140 kVp). Contrast-to-noise ratios and dose-saving potential by spectral shaping were computed for each scanner. Three radiologists independently assessed the image quality of each examination with the intraclass correlation coefficient being computed to measure interrater agreement. Results Regardless of tin prefiltration, PCD-CT with low (171.2 ± 10.3 HU) and intermediate radiation dose (134.7 ± 4.5 HU) provided less image noise than full-dose EID-CT (177.0 ± 14.2 HU; P < 0.001). Targeting matched image noise to 120 kVp EID-CT, mean dose reduction of 79.3% ± 3.9% could be realized in 120 kVp PCD-CT. Subjective image quality of PCD-CT was better than of EID-CT on each dose level ( P < 0.050). While no distinction was found between dose-matched PCD-CT with and without tin prefiltration ( P ≥ 0.928), Sn 150 kVp EID-CT provided better image quality than 120 kVp EID-CT at high and intermediate dose levels ( P > 0.050). The majority of low-dose EID-CT examinations were considered not diagnostic, whereas PCD-CT scans of the same dose level received satisfactory or better ratings. Interrater reliability was excellent (intraclass correlation coefficient 0.903). Conclusions PCD-CT provides superior image quality and significant dose savings compared with EID-CT for ultra-high-resolution examinations of the temporal bone. Aiming for matched image noise, high-voltage scan protocols with tin prefiltration facilitate additional dose saving in EID-CT, whereas superior inherent denoising decreases the dose reduction potential of spectral shaping in PCD-CT.
Abstract Objective Combining fluoroscopy and high-resolution cone-beam CT (CBCT) in a multipurpose scanner without a conventional gantry holds the potential for time-saving in musculoskeletal interventions. This study investigated the performance of gantry-free CBCT arthrography in patients with suspected scapholunate ligament (SLL) tears. Materials and methods Fifty-five patients (29 men, 46.0 ± 15.3 years) who underwent preoperative gantry-free CBCT arthrography between June 2021 and March 2024 were analyzed retrospectively. Three radiologists assessed CBCT arthrograms for tears of the palmar and dorsal SLL segments. Surgical reports served as the reference standard for calculating indicators of diagnostic accuracy. Interreader agreement was tested by computing Krippendorff α . Radiation dose and examination time were recorded. Results Tears of the palmar and dorsal SLL segment were recorded in 25 (45%) and 6 patients (11%), respectively. CBCT arthrography facilitated good sensitivity (range for all readers: 84–92%) and excellent specificity (93–97%) in the assessment of the palmar SLL. For the dorsal SLL, sensitivity (83–100%) and specificity (96–98%) were even higher. Substantial agreement was determined for both the palmar ( α = 0.83, 95% CI: 0.74–0.90) and dorsal SLL (0.84, 0.70–0.95). The mean volume CT dose index for CBCT arthrography was 3.2 ± 1.4 mGy. Not requiring patient repositioning, the median time between fluoroscopy-guided contrast injection and CBCT was 3:05 min (2:31–3:50 min). Conclusion Gantry-free CBCT arthrography allows for excellent accuracy in the preoperative diagnosis of SLL tears with low radiation dose. The ability to alternate between fluoroscopy and CBCT without repositioning facilitates a “one-stop-shop” approach with short examination time. Key Points Question Performing fluoroscopy-guided arthrography and high-resolution cone-beam CT without patient repositioning appears advantageous for the preoperative work-up of distal radius fractures with concomitant scapholunate ligament injuries . Findings Gantry-free cone-beam CT arthrography allowed for short examination times and high diagnostic accuracy for either segment of the scapholunate ligament (89–98% versus surgery) . Clinical relevance Preoperative assessment of scapholunate instability influences treatment since surgeons can reduce radius fractures and perform osteosynthesis via a dorsal portal to simultaneously stabilize the scapholunate compartment or use an additional dorsal access route for ligament suture and transfixation .
