ABSTRACT
The paucity of high-quality medical imaging datasets could be mitigated by machine learning models that generate compositionally diverse images that faithfully represent medical concepts and pathologies. However, large vision-language models are trained on natural images, and the diversity distribution of the generated images substantially differs from that of medical images. Moreover, medical language involves specific and semantically rich vocabulary. Here we describe a domain-adaptation strategy for large vision-language models that overcomes distributional shifts. Specifically, by leveraging publicly available datasets of chest X-ray images and the corresponding radiology reports, we adapted a latent diffusion model pre-trained on pairs of natural images and text descriptors to generate diverse and visually plausible synthetic chest X-ray images (as confirmed by board-certified radiologists) whose appearance can be controlled with free-form medical text prompts. The domain-adaptation strategy for the text-conditioned synthesis of medical images can be used to augment training datasets and is a viable alternative to the sharing of real medical images for model training and fine-tuning.
ABSTRACT
OBJECTIVES: To compare the repeatability and interrelation of various late gadolinium enhancement (LGE) assessment techniques for monitoring fibrotic changes in myocarditis follow-up. MATERIALS AND METHODS: LGE extent change between baseline and 3-month cardiovascular magnetic resonance (CMR) was compared in patients with acute myocarditis using the full width at half maximum (FWHM), gray-scale thresholds at 5 and 6 standard deviations (SD5 and SD6), visual assessment with threshold (VAT) and full manual (FM) techniques. In addition, visual presence score (VPS), visual transmurality score (VTS), and a simplified visual change score (VCS) were assessed. Intraclass-correlation (ICC) was used to evaluate repeatability, and methods were compared using Spearman's correlation. RESULTS: Forty-seven patients (38 male, median age: 27 [IQR: 21; 38] years) were included. LGE extent change differed among quantitative techniques (p < 0.01), with variability in the proportion of patients showing LGE change during follow-up (FWHM: 62%, SD5: 74%, SD6: 66%, VAT: 43%, FM: 60%, VPS: 53%, VTS: 77%, VCS: 89%). Repeatability was highest with FWHM (ICC: 0.97) and lowest with SD5 (ICC: 0.89). Semiquantitative scoring had slightly lower values (VPS ICC: 0.81; VTS ICC: 0.71). VCS repeatability was excellent (ICC: 0.93). VPS and VTS correlated with quantitative techniques, while VCS was positively associated with VPS, VTS, VAT, and FM, but not with FWHM, SD5, and SD6. CONCLUSION: FWHM offers the least observer-dependent LGE follow-up after myocarditis. VPS, VTS, and VCS are practical alternatives, showing reliable correlations with quantitative methods. Classification of patients exhibiting either stable or changing LGE relies on the assessment technique. CLINICAL RELEVANCE STATEMENT: This study shows that LGE monitoring in myocarditis is technique-dependent; the FWHM method yields the most consistent fibrotic tracking results, with scoring-based techniques as reliable alternatives. KEY POINTS: Recognition of fibrotic changes during myocarditis follow-up is significantly influenced by the choice of the quantification technique employed. The FWHM technique ensures highly repeatable tracking of myocarditis-related LGE changes. Segment-based visual scoring and the simplified visual change score offer practical, reproducible alternatives in resource-limited settings.
ABSTRACT
Analyzing vast textual data and summarizing key information from electronic health records imposes a substantial burden on how clinicians allocate their time. Although large language models (LLMs) have shown promise in natural language processing (NLP) tasks, their effectiveness on a diverse range of clinical summarization tasks remains unproven. Here we applied adaptation methods to eight LLMs, spanning four distinct clinical summarization tasks: radiology reports, patient questions, progress notes and doctor-patient dialogue. Quantitative assessments with syntactic, semantic and conceptual NLP metrics reveal trade-offs between models and adaptation methods. A clinical reader study with 10 physicians evaluated summary completeness, correctness and conciseness; in most cases, summaries from our best-adapted LLMs were deemed either equivalent (45%) or superior (36%) compared with summaries from medical experts. The ensuing safety analysis highlights challenges faced by both LLMs and medical experts, as we connect errors to potential medical harm and categorize types of fabricated information. Our research provides evidence of LLMs outperforming medical experts in clinical text summarization across multiple tasks. This suggests that integrating LLMs into clinical workflows could alleviate documentation burden, allowing clinicians to focus more on patient care.
Subject(s)
Documentation , Semantics , Humans , Electronic Health Records , Natural Language Processing , Physician-Patient RelationsABSTRACT
Sifting through vast textual data and summarizing key information from electronic health records (EHR) imposes a substantial burden on how clinicians allocate their time. Although large language models (LLMs) have shown immense promise in natural language processing (NLP) tasks, their efficacy on a diverse range of clinical summarization tasks has not yet been rigorously demonstrated. In this work, we apply domain adaptation methods to eight LLMs, spanning six datasets and four distinct clinical summarization tasks: radiology reports, patient questions, progress notes, and doctor-patient dialogue. Our thorough quantitative assessment reveals trade-offs between models and adaptation methods in addition to instances where recent advances in LLMs may not improve results. Further, in a clinical reader study with ten physicians, we show that summaries from our best-adapted LLMs are preferable to human summaries in terms of completeness and correctness. Our ensuing qualitative analysis highlights challenges faced by both LLMs and human experts. Lastly, we correlate traditional quantitative NLP metrics with reader study scores to enhance our understanding of how these metrics align with physician preferences. Our research marks the first evidence of LLMs outperforming human experts in clinical text summarization across multiple tasks. This implies that integrating LLMs into clinical workflows could alleviate documentation burden, empowering clinicians to focus more on personalized patient care and the inherently human aspects of medicine.
ABSTRACT
OBJECTIVES: The aim of this study was to compare image quality and endoleak detection after endovascular abdominal aortic aneurysm repair between a triphasic computed tomography (CT) with true noncontrast (TNC) and a biphasic CT with virtual noniodine (VNI) images on photon-counting detector CT (PCD-CT). MATERIALS AND METHODS: Adult patients after endovascular abdominal aortic aneurysm repair who received a triphasic examination (TNC, arterial, venous phase) on a PCD-CT between August 2021 and July 2022 were retrospectively included. Endoleak detection was evaluated by 2 blinded radiologists on 2 different readout sets (triphasic CT with TNC-arterial-venous vs biphasic CT with VNI-arterial-venous). Virtual noniodine images were reconstructed from the venous phase. The radiologic report with additional confirmation by an expert reader served as reference standard for endoleak presence. Sensitivity, specificity, and interreader agreement (Krippendorf α) were calculated. Image noise was assessed subjectively in patients using a 5-point scale and objectively calculating the noise power spectrum in a phantom. RESULTS: One hundred ten patients (7 women; age, 76 ± 8 years) with 41 endoleaks were included. Endoleak detection was comparable between both readout sets with a sensitivity and specificity of 0.95/0.84 (TNC) versus 0.95/0.86 (VNI) for reader 1 and 0.88/0.98 (TNC) versus 0.88/0.94 (VNI) for reader 2. Interreader agreement for endoleak detection was substantial (TNC: 0.716, VNI: 0.756). Subjective image noise was comparable between TNC and VNI (4; IQR [4, 5] vs 4; IQR [4, 5], P = 0.44). In the phantom, noise power spectrum peak spatial frequency was similar between TNC and VNI (both f peak = 0.16 mm -1 ). Objective image noise was higher in TNC (12.7 HU) as compared with VNI (11.5 HU). CONCLUSIONS: Endoleak detection and image quality were comparable using VNI images in biphasic CT as compared with TNC images in triphasic CT offering the possibility to reduce scan phases and radiation exposure.
Subject(s)
Aortic Aneurysm, Abdominal , Endoleak , Adult , Humans , Female , Aged , Aged, 80 and over , Endoleak/diagnostic imaging , Computed Tomography Angiography/methods , Endovascular Aneurysm Repair , Retrospective Studies , Tomography, X-Ray Computed/methods , Aortic Aneurysm, Abdominal/diagnostic imaging , Aortic Aneurysm, Abdominal/surgeryABSTRACT
PURPOSE: Numerous electrocardiogram (ECG) abnormalities and late gadolinium enhancement (LGE) in cardiac magnetic resonance imaging (CMR) have been related to poor prognosis in acute myocarditis. We evaluated whether ECG parameters are associated with the distribution and dynamic of LGE along the course of myocarditis. METHODS: Fifty-one patients with CMR confirmed acute myocarditis were included who underwent CMR with LGE and 12-lead ECG at baseline and 3-month follow-up at our institution. The association between the presence, regional distribution and change of ECG parameters and LGE was investigated using linear regression analysis. LGE was quantified as visual presence score (VPS) and visual transmurality score (VTS). RESULTS: Among many ECG parameters only > 1 mm ST-elevation (STE) was associated with VPS and VTS at baseline (ß = 3.08 [95%CI: 1.75; 4.41], p = < 0.001 and ß = 5.40 [95%CI: 1.92; 8.88], p = 0.004; respectively). STE was most frequent in lateral and inferior ECG-leads (48% and 31%) and it was associated with VPS and VTS in these localizations (p < 0.05 all), however no association between anterior-septal STE and LGE could be confirmed. At follow-up the regression of STE was associated with the regression of VPS and VTS in univariate analysis (ß=-1.49 [95%CI: -2.41; -0.57], p = 0.003 and ß=-4.87 [95%CI: -7.18; -2.56], p = 0.001, respectively), which remained significant for VTS using a multivariate model (ß=-2.39 [95%CI: -3.32; -0.47], p = 0.019). CONCLUSION: Although we demonstrated some promising associations between STE and LGE, the usability of ECG to estimate the territorial involvement and dynamical changes of LGE along the course of myocarditis is generally limited and cardiac magnetic resonance should be considered for this purpose.
Subject(s)
Myocarditis , Humans , Myocarditis/pathology , Contrast Media , Gadolinium , Magnetic Resonance Imaging, Cine/methods , Predictive Value of Tests , Electrocardiography/methods , Arrhythmias, CardiacABSTRACT
OBJECTIVES: In patients of advanced age, the feasibility of myocardial ischemia testing might be limited by age-related comorbidities and falling compliance abilities. Therefore, we aimed to test the accuracy of 3D cardiac magnetic resonance (CMR) stress perfusion in the elderly population as compared to reference standard fractional flow reserve (FFR). METHODS: Fifty-six patients at age 75 years or older (mean age 79 ± 4 years, 35 male) underwent 3D CMR perfusion imaging and invasive coronary angiography with FFR in 5 centers using the same study protocol. The diagnostic accuracy of CMR was compared to a control group of 360 patients aged below 75 years (mean age 61 ± 9 years, 262 male). The percentage of myocardial ischemic burden (MIB) relative to myocardial scar burden was further analyzed using semi-automated software. RESULTS: Sensitivity, specificity, and positive and negative predictive values of 3D perfusion CMR deemed similar for both age groups in the detection of hemodynamically relevant (FFR < 0.8) stenosis (≥ 75 years: 86%, 83%, 92%, and 75%; < 75 years: 87%, 80%, 82%, and 85%; p > 0.05 all). While MIB was larger in the elderly patients (15% ± 17% vs. 9% ± 13%), the diagnostic accuracy of 3D CMR perfusion was high in both elderly and non-elderly populations to predict pathological FFR (AUC: 0.906 and 0.866). CONCLUSIONS: 3D CMR perfusion has excellent diagnostic accuracy for the detection of hemodynamically relevant coronary stenosis, independent of patient age. KEY POINTS: ⢠The increasing prevalence of coronary artery disease in elderly populations is accompanied with a larger ischemic burden of the myocardium as compared to younger individuals. ⢠3D cardiac magnetic resonance perfusion imaging predicts pathological fractional flow reserve in elderly patients aged ≥ 75 years with high diagnostic accuracy. ⢠Ischemia testing with 3D CMR perfusion imaging has similarly high accuracy in the elderly as in younger patients and it might be particularly useful when other non-invasive techniques are limited by aging-related comorbidities and falling compliance abilities.
Subject(s)
Coronary Artery Disease , Coronary Stenosis , Fractional Flow Reserve, Myocardial , Myocardial Perfusion Imaging , Humans , Male , Aged , Aged, 80 and over , Middle Aged , Coronary Artery Disease/diagnosis , Myocardial Perfusion Imaging/methods , Severity of Illness Index , Coronary Angiography/methods , Predictive Value of Tests , Perfusion , Magnetic Resonance SpectroscopyABSTRACT
Purpose: To assess the impact of virtual-monoenergetic-image (VMI) energies on the diagnosis of pulmonary embolism (PE) in photon-counting-detector computed-tomography (PCD-CT). Methods: Eighty patients (median age 60.4 years) with suspected PE were retrospectively included. Scans were performed on PCD-CT in the multi-energy mode at 120 kV. VMIs from 40−70 keV in 10 keV intervals were reconstructed. CT-attenuation was measured in the pulmonary trunk and the main branches of the pulmonary artery. Signal-to-noise (SNR) ratio was calculated. Two radiologists evaluated subjective-image-quality (noise, vessel-attenuation and sharpness; five-point-Likert-scale, non-diagnostic−excellent), the presence of hardening artefacts and presence/visibility of PE. Results: Signal was highest at the lowest evaluated VMI (40 keV; 1053.50 HU); image noise was lowest at the highest VMI (70 keV; 15.60 HU). Highest SNR was achieved at the lowest VMI (p < 0.05). Inter-reader-agreement for subjective analysis was fair to excellent (k = 0.373−1.000; p < 0.001). Scores for vessel-attenuation and sharpness were highest at 40 keV (both:5, range 4/3−5; k = 1.000); scores for image-noise were highest at 70 keV (4, range 3−5). The highest number of hardening artifacts were reported at 40 keV (n = 22; 28%). PE-visualization was rated best at 50 keV (4.7; range 4−5) and decreased with increasing VMI-energy (r = −0.558; p < 0.001). Conclusions: While SNR was best at 40 keV, subjective PE visibility was rated highest at 50 keV, potentially owing to the lower image noise and hardening artefacts.
ABSTRACT
PURPOSE: To find the best level to measure aortic flow for quantification of aortic regurgitation (AR) in 4D flow CMR. METHODS: In 27 congenital heart disease patients with AR (67% male, 31 ± 16 years) two blinded observers measured antegrade, retrograde, net aortic flow volumes and regurgitant fractions at 6 levels in 4D flow: (1) below the aortic valve (AV), (2) at the AV, (3) at the aortic sinus, (4) at the sinotubular junction, (5) at the level of the pulmonary arteries (PA) and (6) below the brachiocephalic trunk. 2D phase contrast (2DPC) sequences were acquired at the level of PA. All patients received prior transthoracic echocardiography (TTE) with AR severity grading according to a recommended multiparametric approach. RESULTS: After assigning 2DPC measurements into AR grading, agreement between TTE AR grading and 2DPC was good (κ = 0.88). In 4D flow, antegrade flow was similar between the six levels (p = 0.87). Net flow was higher at level 1-2 than at levels 3-6 (p < 0.05). Retrograde flow and regurgitant fraction at level 1-2 were lower compared to levels 3-6 (p < 0.05). Reproducibility (inter-reader agreement: ICC 0.993, 95% CI 0.986-0.99; intra-reader agreement: ICC 0.982, 95%CI 0.943-0.994) as well as measurement agreement between 4D flow and 2DPC (ICC 0.994; 95%CI 0.989 - 0.998) was best at the level of PA. CONCLUSION: For estimating severity of AR in 4D flow, best reproducibility along with best agreement with 2DPC measurements can be expected at the level of PA. Measurements at AV or below AV might underestimate AR.
Subject(s)
Aortic Valve Insufficiency , Aortic Valve/diagnostic imaging , Aortic Valve Insufficiency/diagnostic imaging , Female , Humans , Magnetic Resonance Imaging, Cine , Magnetic Resonance Spectroscopy , Male , Reproducibility of Results , Severity of Illness IndexABSTRACT
PURPOSE: To propose respiratory motion-informed locally low-rank reconstruction (MI-LLR) for robust free-breathing single-bolus quantitative 3D myocardial perfusion CMR imaging. Simulation and in-vivo results are compared to locally low-rank (LLR) and compressed sensing reconstructions (CS) for reference. METHODS: Data were acquired using a 3D Cartesian pseudo-spiral in-out k-t undersampling scheme (R = 10) and reconstructed using MI-LLR, which encompasses two stages. In the first stage, approximate displacement fields are derived from an initial LLR reconstruction to feed a motion-compensated reference system to a second reconstruction stage, which reduces the rank of the inverse problem. For comparison, data were also reconstructed with LLR and frame-by-frame CS using wavelets as sparsifying transform ( â1$$ {\ell}_1 $$ -wavelet). Reconstruction accuracy relative to ground truth was assessed using synthetic data for realistic ranges of breathing motion, heart rates, and SNRs. In-vivo experiments were conducted in healthy subjects at rest and during adenosine stress. Myocardial blood flow (MBF) maps were derived using a Fermi model. RESULTS: Improved uniformity of MBF maps with reduced local variations was achieved with MI-LLR. For rest and stress, intra-volunteer variation of absolute and relative MBF was lower in MI-LLR (±0.17 mL/g/min [26%] and ±1.07 mL/g/min [33%]) versus LLR (±0.19 mL/g/min [28%] and ±1.22 mL/g/min [36%]) and versus â1$$ {\ell}_1 $$ -wavelet (±1.17 mL/g/min [113%] and ±6.87 mL/g/min [115%]). At rest, intra-subject MBF variation was reduced significantly with MI-LLR. CONCLUSION: The combination of pseudo-spiral Cartesian undersampling and dual-stage MI-LLR reconstruction improves free-breathing quantitative 3D myocardial perfusion CMR imaging under rest and stress condition.
Subject(s)
Myocardial Perfusion Imaging , Adenosine , Coronary Circulation , Humans , Magnetic Resonance Imaging/methods , Motion , Myocardial Perfusion Imaging/methods , RespirationABSTRACT
BACKGROUND: The purpose of the study was to investigate feasibility of infarct detection in segmental strain derived from non-contrast cardiac magnetic resonance (CMR) cine sequences in patients with acute myocardial infarction (AMI) and in follow-up (FU) exams. METHODS: 57 patients with AMI (mean age 61 ± 12 years, CMR 2.8 ± 2 days after infarction) were retrospectively included, FU exams were available in 32 patients (35 ± 14 days after first CMR). 43 patients with normal CMR (54 ± 11 years) served as controls. Dedicated software (Segment CMR, Medviso) was used to calculate global and segmental strain derived from cine sequences. Cine short axis stacks and segmental circumferential strain calculations of every patient and control were presented to two blinded readers in random order, who were advised to identify potentially infarcted segments, blinded to LGE and clinical information. RESULTS: Impaired global strain was measured in AMI patients compared to controls (global peak circumferential strain [GPCS] p = 0.01; global peak longitudinal strain [GPLS] p = 0.04; global peak radial strain [GPRS] p = 0.01). In both imaging time points, mean segmental peak circumferential strain [SPCS] was impaired in infarcted tissue compared to remote segments (AMI: p = 0.03, FU: p = 0.02). SPCS values in infarcted segments were similar between AMI and FU (p = 0.8). In SPCS calculations, 141 from 189 acutely infarcted segments were accurately detected (74.6%), visual evaluation of correlating cine images detected 43.4% infarcts. In FU, 80% infarcted segments (91/114 segments) were detected in SPCS and 51.8% by visual evaluation of correlating short axis cine images (p = 0.01). CONCLUSION: Segmental circumferential strain derived from routinely acquired native cine sequences detects nearly 75% of acute infarcts and 80% of infarcts in subacute follow-up CMR, significantly more than visual evaluation of correlating cine images alone. Acute infarcts may display only subtle impairment of wall motion and no obvious wall thinning, thus SPCS calculation might be helpful for scar detection in patients with acute infarcts, when LGE images are not available.
Subject(s)
Cicatrix , Myocardial Infarction , Aged , Cicatrix/diagnostic imaging , Cicatrix/etiology , Cicatrix/pathology , Follow-Up Studies , Humans , Magnetic Resonance Imaging, Cine/methods , Magnetic Resonance Spectroscopy , Middle Aged , Myocardial Infarction/diagnostic imaging , Myocardial Infarction/pathology , Myocardium/pathology , Retrospective Studies , Ventricular Function, LeftABSTRACT
OBJECTIVES: Although cardiovascular magnetic resonance (CMR) is increasingly used to diagnose pericardial inflammation, imaging can still be challenging using conventional CMR techniques. Parametric mapping (T1/T2 mapping) techniques have emerged as novel methods to quantify focal and global changes of the myocardium without contrast agent. The aim of the present study was to implement parametric mapping to facilitate diagnostic decision-making in pericardial inflammation. METHODS: Twenty patients with pericardial inflammation underwent CMR (1.5T system) including T1-weighted/T2-weighted imaging, T1/T2 mapping and late gadolinium enhancement. T1/T2 mapping was performed in end-diastole covering three short-axis slices. Diagnosis of pericardial inflammation was made according to recent guidelines. T1/T2 measurements were pursued by manually drawing regions of interest (ROIs) in the thickened, diseased pericardium carefully avoiding contamination by other cardiac structures. Parametric values were correlated to further markers of pericardial inflammation, such as pericardial thickening and inflammatory parameters. RESULTS: On average, the pericardium displayed a thickness of 4.8±1.0 mm. Mean T1 value was 1363.0±227.1 ms and T2 value was 123.3±52.6 ms, which were above patient's myocardial values (myocardial T1: 998.7±81.0 ms, p<0.001, median 1014.46 ms; T2: 68.0±28.9 m,p<0.001) and the values of a group of four patients with chronic pericarditis (T1: 953.0±16.7 ms; T2: 63.2±10.1 ms). T1 and T2 showed a correlation to the extent of the thickened pericardium (R=0.64, p=0.002 for T1, R=0.72, p=0.005 for T2). There was no correlation of pericardial T1/T2 to blood markers of inflammation, myocardial injury (C reactive protein, troponin, creatine kinase) or further CMR parameters. CONCLUSIONS: In patients with pericardial inflammation, parametric mapping showed elevated T1 and T2 values. Parametric mapping may help to facilitate diagnosis of pericardial inflammation if conventional parameters such as pericardial hyperintensity in T1-weighted or T2-weighted imaging or contrast agent uptake are heterogeneous.
Subject(s)
Contrast Media , Gadolinium , Humans , Inflammation/diagnostic imaging , Magnetic Resonance Spectroscopy , Pericardium/diagnostic imaging , Predictive Value of TestsABSTRACT
OBJECTIVES: Late gadolinium enhancement with fixed short inversion time (LGEshort) provides excellent tissue contrast with dark scar and bright blood pool and does not need prior myocardial nulling. We hypothesize better visibility of ischemic scars and equal visibility of non-ischemic LGE in LGEshort compared to clinically established LGE (LGEstandard). METHODS: LGEshort and LGEstandard were retrospectively evaluated in 179 patients (3043 segments) with suspected or known coronary artery disease by four blinded readers (reader A: most experienced - D: least experienced). The amount of ischemic and non-ischemic LGE as well as visibility (4: very good - 1: poor) of ischemic LGE was visually assessed. RESULTS: All readers detected more infarcted segments in LGEshort compared to LGEstandard (378 segments reported as infarcted; A:p = 0.5, B:p = 0.8, C,D:p = 0.03). Scar visibility was scored higher in LGEshort by all readers (A,B:p = 0.03; C,D:p = 0.02), especially for subendocardial infarcts (A,B:p = 0.04, C,D:p = 0.02). Less experienced readers detected significantly more infarcted papillary muscles (C:p = 0.02, D:p = 0.03) in a shorter reading time in LGEshort (C:p = 0.04, D:p = 0.02). Non-ischemic LGE was equally visible in both sequences (A:p = 0.9, B:p = 0.8, C,D:p = 0.6). CONCLUSIONS: LGEshort detects more ischemic LGE with improved scar visibility compared to LGEstandard, independent of experience level. The visibility of non-ischemic LGE is equivalent to LGEstandard. Less experienced readers can diagnose ischemic and non-ischemic LGE faster in LGEshort. ADVANCES IN KNOWLEDGE: LGEshort with its maximal operational simplicity can be used for visualization of all types of fibrosis - ischemic and non-ischemic - instead of LGEstandard, independent of experience level.
Subject(s)
Cardiomyopathies , Gadolinium , Cicatrix/diagnostic imaging , Contrast Media , Fibrosis , Humans , Image Enhancement , Ischemia , Magnetic Resonance Imaging/methods , Magnetic Resonance Spectroscopy , Myocardium/pathology , Predictive Value of Tests , Retrospective StudiesABSTRACT
Purpose: To compare three-dimensional (3D) whole-heart MRI with isotropic submillimeter resolution with standard two-dimensional (2D) cine MRI in measuring the bilayered myocardium in left ventricular noncompaction (LVNC). Materials and Methods: Twenty-four patients with LVNC (mean age, 42 years ± 16 [SD]) were retrospectively enrolled between October 2011 and July 2020. Compacted myocardium (CM) and noncompacted myocardium (NCM) were measured in long axis (Petersen approach) and short axis (Jacquier approach) at 3D whole-heart and 2D cine MRI by two independent readers. Image quality (1 = excellent, 2 = adequate, 3 = nondiagnostic), considering discrimination between NCM and CM and CM and adjacent tissue, was evaluated. Pearson, Spearman, and intraclass correlation tests were used as statistical tests. Results: In long-axis measurements, the correlation between both sequences was moderate to strong for CM (Pearson, 0.66-0.79; Spearman, 0.61-0.68) and strong to very strong for NCM (Pearson, 0.90-0.97; Spearman, 0.77-0.91). Intraclass correlation coefficient (ICC) in 3D whole-heart MRI was 0.90 (95% CI: 0.78, 0.95) for CM and 0.94 (95% CI: 0.84, 0.97) for NCM, while ICC in 2D cine MRI was 0.77 (95% CI: 0.55, 0.89) for CM and 0.87 (95% CI: 0.72, 0.94) for NCM. Short-axis CM and NCM measurements had a strong to very strong correlation between both sequences (Pearson, 0.86-0.98; Spearman, 0.82-0.98). ICC in 3D whole-heart MRI was 0.96 (95% CI: 0.94, 0.99) for CM and 0.98 (95% CI: 0.97, 0.99) for NCM, while ICC in 2D cine MRI was 0.82 (95% CI: 0.63, 0.92) for CM and 0.87 (95% CI: 0.72, 0.94) for NCM. 3D whole-heart MRI demonstrated higher image quality than did 2D cine MRI (P < .001). Conclusion: 3D whole-heart MRI revealed higher image quality, with better structure discrimination and interobserver agreement in LVNC measurements, compared with standard 2D cine images.Keywords: MR Imaging, Cardiac, Cardiovascular Magnetic Resonance, Left Ventricular Noncompaction, Free-breathing Imaging Technique Supplemental material is available for this article. © RSNA, 2022See also the commentary by Jensen and Petersen in this issue.
ABSTRACT
OBJECTIVES: Computed tomography angiography (CTA) is broadly used for long-term follow-up of graft patency after coronary artery bypass graft surgery (CABG). However, its clinical value in the early postoperative setting has not been established yet. We evaluated the benefit of adding CTA to the routine clinical work-up after CABG on patient management. METHODS: A total of 305 consecutive patients (269 males, median age 68 years) underwent CABG and postoperative CTA with a median of 6 days after surgery. Graft patency and additional imaging findings were assessed and their influence on diagnosis and clinical management was evaluated. RESULTS: Graft occlusion or high-grade stenosis was found in 15% of the patients. Additional findings were reported in 44% of the patients, including pericardial (2%) and pleural effusion (27%), large pneumothorax (11%), pulmonary infection (4%), cardiac or vascular thrombus (2%), pulmonary embolism (2%), sternal dehiscence (1%) and additional incidental findings requiring follow-up (6%). CT findings initiated new diagnostic and/or therapeutic measures in 15% of the patients, 47% of those with diseased grafts and 19% of patients with non-graft-related findings. No adverse events related to CTA were documented. CONCLUSIONS: Early routine postoperative assessment of CABG with CTA reveals both cardiac and non-cardiac findings with a high frequency, affecting clinical management in a substantial proportion of patients.
Subject(s)
Computed Tomography Angiography , Graft Occlusion, Vascular , Aged , Coronary Angiography/methods , Coronary Artery Bypass/adverse effects , Coronary Artery Bypass/methods , Humans , Male , Tomography, X-Ray Computed/methods , Vascular PatencyABSTRACT
Aim: This study investigated a patient-specific approach of using cardiac magnetic resonance (CMR) feature tracking for scar detection in a heterogenous patient group with chronic ischemic and non-ischemic heart disease. Methods: CMR exams of 89 patients with concomitant chronic ischemic and non-ischemic heart disease (IHD+) as well as 65 patients with ischemic scars only (IHD) were retrospectively evaluated. In all patients, global (GCS) and segmental circumferential strain (SCS) was derived from native cine images using a dedicated software (Segment CMR, Medviso). After calculation of patient-specific median GCS (GCSmedian), segmental values from GCSmedian percentage plots were correlated with corresponding myocardial segments in late gadolinium enhancement (LGE). Results: Overall GCS ranged between -3.5% to -19.8% and average GCS was lower in IHD+ than in IHD (p <0.05). In IHD, 19% of all myocardial segments were infarcted, in IHD+ 16.6%. Additionally, non-ischemic LGE was present in 6.7% of segments in IHD+. Correlation of GCSmedian percentage plots with corresponding LGE showed that presence of ischemic scar tissue in a myocardial segment was very likely below a cut-off of 39.5% GCSmedian (87.5% sensitivity, 86.3% specificity, AUC 0.907, 95% CI 0.875-0.938, p < 0.05). Conclusion: In patient-specific GCSmedian percentage plots calculated from native cine images, ischemic scar tissue can be suspected in myocardial segments below the threshold of 40% GCSmedian (sensitivity 88%, specificity 86%), even in a heterogenous patient cohort with ischemic and non-ischemic heart disease.
ABSTRACT
OBJECTIVE: The aim of this study was to evaluate the image quality (IQ) and performance of an artificial intelligence (AI)-based computer-aided detection (CAD) system in photon-counting detector computed tomography (PCD-CT) for pulmonary nodule evaluation at different low-dose levels. MATERIALS AND METHODS: An anthropomorphic chest-phantom containing 14 pulmonary nodules of different sizes (range, 3-12 mm) was imaged on a PCD-CT and on a conventional energy-integrating detector CT (EID-CT). Scans were performed with each of the 3 vendor-specific scanning modes (QuantumPlus [Q+], Quantum [Q], and High Resolution [HR]) at decreasing matched radiation dose levels (volume computed tomography dose index ranging from 1.79 to 0.31 mGy) by adapting IQ levels from 30 to 5. Image noise was measured manually in the chest wall at 8 different locations. Subjective IQ was evaluated by 2 readers in consensus. Nodule detection and volumetry were performed using a commercially available AI-CAD system. RESULTS: Subjective IQ was superior in PCD-CT compared with EID-CT (P < 0.001), and objective image noise was similar in the Q+ and Q-mode (P > 0.05) and superior in the HR-mode (PCD 55.8 ± 11.7 HU vs EID 74.8 ± 5.4 HU; P = 0.01). High resolution showed the lowest image noise values among PCD modes (P = 0.01). Overall, the AI-CAD system delivered comparable results for lung nodule detection and volumetry between PCD- and dose-matched EID-CT (P = 0.08-1.00), with a mean sensitivity of 95% for PCD-CT and of 86% for dose-matched EID-CT in the lowest evaluated dose level (IQ5). Q+ and Q-mode showed higher false-positive rates than EID-CT at lower-dose levels (IQ10 and IQ5). The HR-mode showed a sensitivity of 100% with a false-positive rate of 1 even at the lowest evaluated dose level (IQ5; CDTIvol, 0.41 mGy). CONCLUSIONS: Photon-counting detector CT was superior to dose-matched EID-CT in subjective IQ while showing comparable to lower objective image noise. Fully automatized AI-aided nodule detection and volumetry are feasible in PCD-CT, but attention has to be paid to false-positive findings.
Subject(s)
Artificial Intelligence , Photons , Computers , Phantoms, Imaging , Tomography, X-Ray Computed/methodsABSTRACT
BACKGROUND: We examined the dynamic response of the myocardium to infarction in a longitudinal porcine study using relaxometry, functional as well as diffusion cardiovascular magnetic resonance (CMR). We sought to compare non contrast CMR methods like relaxometry and in-vivo diffusion to contrast enhanced imaging and investigate the link of microstructural and functional changes in the acute and chronically infarcted heart. METHODS: CMR was performed on five myocardial infarction pigs and four healthy controls. In the infarction group, measurements were obtained 2 weeks before 90 min occlusion of the left circumflex artery, 6 days after ischemia and at 5 as well as 9 weeks as chronic follow-up. The timing of measurements was replicated in the control cohort. Imaging consisted of functional cine imaging, 3D tagging, T2 mapping, native as well as gadolinium enhanced T1 mapping, cardiac diffusion tensor imaging, and late gadolinium enhancement imaging. RESULTS: Native T1, extracellular volume (ECV) and mean diffusivity (MD) were significantly elevated in the infarcted region while fractional anisotropy (FA) was significantly reduced. During the transition from acute to chronic stages, native T1 presented minor changes (< 3%). ECV as well as MD increased from acute to the chronic stages compared to baseline: ECV: 125 ± 24% (day 6) 157 ± 24% (week 5) 146 ± 60% (week 9), MD: 17 ± 7% (day 6) 33 ± 14% (week 5) 29 ± 15% (week 9) and FA was further reduced: - 31 ± 10% (day 6) - 38 ± 8% (week 5) - 36 ± 14% (week 9). T2 as marker for myocardial edema was significantly increased in the ischemic area only during the acute stage (83 ± 3 ms infarction vs. 58 ± 2 ms control p < 0.001 and 61 ± 2 ms in the remote area p < 0.001). The analysis of functional imaging revealed reduced left ventricular ejection fraction, global longitudinal strain and torsion in the infarct group. At the same time the transmural helix angle (HA) gradient was steeper in the chronic follow-up and a correlation between longitudinal strain and transmural HA gradient was detected (r = 0.59 with p < 0.05). Comparing non-gadolinium enhanced data T2 mapping showed the largest relative change between infarct and remote during the acute stage (+ 33 ± 4% day 6, with p = 0.013 T2 vs. MD, p = 0.009 T2 vs. FA and p = 0.01 T2 vs. T1) while FA exhibited the largest relative change between infarct and remote during the chronic follow-up (+ 31 ± 2% week 5, with p = N.S. FA vs. MD, p = 0.03 FA vs. T2 and p = 0.003 FA vs. T1). Overall, diffusion parameters provided a higher contrast (> 23% for MD and > 27% for FA) during follow-up compared to relaxometry (T1 17-18%/T2 10-20%). CONCLUSION: During chronic follow-up after myocardial infarction, cardiac diffusion tensor imaging provides a higher sensitivity for mapping microstructural alterations when compared to non-contrast enhanced relaxometry with the added benefit of providing directional tensor information to assess remodelling of myocyte aggregate orientations, which cannot be otherwise assessed.