Dataset Information

Atri-U: assisted image analysis in routine cardiovascular magnetic resonance volumetry of the left atrium

ABSTRACT:

Background

Artificial intelligence can assist in cardiac image interpretation. Here, we achieved a substantial reduction in time required to read a cardiovascular magnetic resonance (CMR) study to estimate left atrial volume without compromising accuracy or reliability. Rather than deploying a fully automatic black-box, we propose to incorporate the automated LA volumetry into a human-centric interactive image-analysis process.

Methods and results

Atri-U, an automated data analysis pipeline for long-axis cardiac cine images, computes the atrial volume by: (i) detecting the end-systolic frame, (ii) outlining the endocardial borders of the LA, (iii) localizing the mitral annular hinge points and constructing the longitudinal atrial diameters, equivalent to the usual workup done by clinicians. In every step human interaction is possible, such that the results provided by the algorithm can be accepted, corrected, or re-done from scratch. Atri-U was trained and evaluated retrospectively on a sample of 300 patients and then applied to a consecutive clinical sample of 150 patients with various heart conditions. The agreement of the indexed LA volume between Atri-U and two experts was similar to the inter-rater agreement between clinicians (average overestimation of 0.8 mL/m2 with upper and lower limits of agreement of − 7.5 and 5.8 mL/m2, respectively). An expert cardiologist blinded to the origin of the annotations rated the outputs produced by Atri-U as acceptable in 97% of cases for step (i), 94% for step (ii) and 95% for step (iii), which was slightly lower than the acceptance rate of the outputs produced by a human expert radiologist in the same cases (92%, 100% and 100%, respectively). The assistance of Atri-U lead to an expected reduction in reading time of 66%—from 105 to 34 s, in our in-house clinical setting.

Conclusions

Our proposal enables automated calculation of the maximum LA volume approaching human accuracy and precision. The optional user interaction is possible at each processing step. As such, the assisted process sped up the routine CMR workflow by providing accurate, precise, and validated measurement results.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12968-021-00791-8.

SUBMITTER: Anastasopoulos C

PROVIDER: S-EPMC8582149 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Cardiovascular resonance (CMR) imaging is a standard imaging modality for assessing cardiovascular diseases (CVDs), the leading cause of death globally. CMR enables accurate quantification of the cardiac chamber volume, ejection fraction and myocardial mass, providing information for diagnosis and monitoring of CVDs. However, for years, clinicians have been relying on manual approaches for CMR image analysis, which is time consuming and prone to subjective errors. It is a major clinical challenge to automatically derive quantitative and clinically relevant information from CMR images.Deep neural networks have shown a great potential in image pattern recognition and segmentation for a variety of tasks. Here we demonstrate an automated analysis method for CMR images, which is based on a fully convolutional network (FCN). The network is trained and evaluated on a large-scale dataset from the UK Biobank, consisting of 4,875 subjects with 93,500 pixelwise annotated images. The performance of the method has been evaluated using a number of technical metrics, including the Dice metric, mean contour distance and Hausdorff distance, as well as clinically relevant measures, including left ventricle (LV) end-diastolic volume (LVEDV) and end-systolic volume (LVESV), LV mass (LVM); right ventricle (RV) end-diastolic volume (RVEDV) and end-systolic volume (RVESV).By combining FCN with a large-scale annotated dataset, the proposed automated method achieves a high performance in segmenting the LV and RV on short-axis CMR images and the left atrium (LA) and right atrium (RA) on long-axis CMR images. On a short-axis image test set of 600 subjects, it achieves an average Dice metric of 0.94 for the LV cavity, 0.88 for the LV myocardium and 0.90 for the RV cavity. The mean absolute difference between automated measurement and manual measurement is 6.1 mL for LVEDV, 5.3 mL for LVESV, 6.9 gram for LVM, 8.5 mL for RVEDV and 7.2 mL for RVESV. On long-axis image test sets, the average Dice metric is 0.93 for the LA cavity (2-chamber view), 0.95 for the LA cavity (4-chamber view) and 0.96 for the RA cavity (4-chamber view). The performance is comparable to human inter-observer variability.We show that an automated method achieves a performance on par with human experts in analysing CMR images and deriving clinically relevant measures.

Project description:BackgroundAssessing the structure and function of left atrium (LA) is crucial in hypertrophic obstructive cardiomyopathy (HOCM) because LA remodeling correlates with atrial fibrillation. However, few studies have investigated the potential effect of myomectomy on LA phasic remodeling in HOCM after myectomy using cardiovascular magnetic resonance (CMR) feature tracking (FT). This study aims to evaluate the LA structural and functional remodeling with HOCM after myectomy by CMR-FT and to further investigate the determinants of LA reverse remodeling.MethodsIn this single-center study, we retrospectively studied 88 patients with HOCM who received CMR before and after myectomy between January 2011 and June 2021. Preoperative and postoperative LA parameters derived from CMR-FT were compared, including LA reservoir function (total ejection fraction [EF], total strain [εs], peak positive strain rate [SRs]), conduit function (passive EF, passive strain [εe], peak early negative strain rate [SRe]) and booster function (booster EF, active strain [εa], late peak negative strain rate [SRa]). Eighty-six healthy participants were collected for comparison. Univariate and multivariate linear regression identified variables associated with the rate of change of εa.ResultsCompared with preoperative parameters, LA reservoir function (total EF, εs, SRs), booster function (booster EF, εa, SRa), and SRe were significantly improved after myectomy (all P < 0.05), while no significant differences were observed in passive EF and εe. Postoperative patients with HOCM still had larger LA and worse LA function than healthy controls (all P < 0.05). After analyzing the rates of change in LA parameters, LA boost function, especially εa, showed the most dramatic improvement beyond the improvements in reservoir function, conduit function, and volume. In multivariable regression analysis, minimum LA volume index (adjusted β = - 0.39, P < 0.001) and Δleft ventricular outflow tract (LVOT) pressure gradient (adjusted β = - 0.29, P = 0.003) were significantly related to the rate of change of εa.ConclusionsPatients with HOCM after septal myectomy showed LA reverse remodeling with a reduction in LA size and restoration in LA reservoir and booster function but unchanged LA conduit function. Among volumetric and functional changes, booster function had the greatest improvement postoperatively. Besides, preoperative LAVmin index and ΔLVOT might be potential factors associated with the degree of improvement in εa.

Project description:BackgroundAccurate measurement of left atrial (LA) volumes is needed in cardiac diagnostics and the follow up of heart and valvular diseases. Geometrical assumptions with 2D methods for LA volume estimation contribute to volume misestimation. In this study, we test agreement of 3D and 2D methods of LA volume detection and explore contribution of 3D LA axis orientation and LA shape in introducing error in 2D methods by cardiovascular magnetic resonance imaging.Methods30 patients with prior first-ever ischemic stroke and no known heart disease, and 30 healthy controls were enrolled (age 18-49) in a substudy of a prospective case-control study. All study subjects underwent cardiac magnetic resonance imaging and were pooled for this methodological study. LA volumes were calculated by biplane area-length method from both conventional long axis (LAVAL-LV) and LA long axis-oriented images (LAVAL-LA) and were compared to 3D segmented LA volume (LAVSAX) to assess accuracy of volume detection. 3D orientation of LA long axis to left ventricular (LV) long axis and to four-chamber plane were determined, and LA 3D sphericity indices were calculated to assess sources of error in LA volume calculation. Shapiro-Wilk test, Bland-Altman analysis, intraclass and Pearson correlation, and Spearman's rho were used for statistical analysis.ResultsBiases were - 9.9 mL (- 12.5 to - 7.2) for LAVAL-LV and 13.4 (10.0-16.9) for LAVAL-LA [mean difference to LAVSAX (95% confidence interval)]. End-diastolic LA long axis 3D deviation angle to LV long axis was 28.3 ± 6.2° [mean ± SD] and LA long axis 3D rotation angle to four-chamber plane 20.5 ± 18.0°. 3D orientation of LA axis or 3D sphericity were not correlated to error in LA volume calculation.ConclusionsCalculated LA volume accuracy did not improve by using LA long axis-oriented images for volume calculation in comparison to conventional method. We present novel data on LA axis orientation and a novel metric of LA sphericity and conclude that these measures cannot be utilized to assess error in LA volume calculation.Trial registrationMain study Searching for Explanations for Cryptogenic Stroke in the Young: Revealing the Etiology, Triggers, and Outcome (SECRETO; NCT01934725) has been registered previously.

Project description:To examine relationships between severity of echocardiography (echo) -evidenced diastolic dysfunction (DD) and volumetric filling by automated processing of routine cine cardiovascular magnetic resonance (CMR).Cine-CMR provides high-resolution assessment of left ventricular (LV) chamber volumes. Automated segmentation (LV-METRIC) yields LV filling curves by segmenting all short-axis images across all temporal phases. This study used cine-CMR to assess filling changes that occur with progressive DD.115 post-MI patients underwent CMR and echo within 1 day. LV-METRIC yielded multiple diastolic indices - E:A ratio, peak filling rate (PFR), time to peak filling rate (TPFR), and diastolic volume recovery (DVR80 - proportion of diastole required to recover 80% stroke volume). Echo was the reference for DD.LV-METRIC successfully generated LV filling curves in all patients. CMR indices were reproducible (< or = 1% inter-reader differences) and required minimal processing time (175 +/- 34 images/exam, 2:09 +/- 0:51 minutes). CMR E:A ratio decreased with grade 1 and increased with grades 2-3 DD. Diastolic filling intervals, measured by DVR80 or TPFR, prolonged with grade 1 and shortened with grade 3 DD, paralleling echo deceleration time (p < 0.001). PFR by CMR increased with DD grade, similar to E/e' (p < 0.001). Prolonged DVR80 identified 71% of patients with echo-evidenced grade 1 but no patients with grade 3 DD, and stroke-volume adjusted PFR identified 67% with grade 3 but none with grade 1 DD (matched specificity = 83%). The combination of DVR80 and PFR identified 53% of patients with grade 2 DD. Prolonged DVR80 was associated with grade 1 (OR 2.79, CI 1.65-4.05, p = 0.001) with a similar trend for grade 2 (OR 1.35, CI 0.98-1.74, p = 0.06), whereas high PFR was associated with grade 3 (OR 1.14, CI 1.02-1.25, p = 0.02) DD.Automated cine-CMR segmentation can discern LV filling changes that occur with increasing severity of echo-evidenced DD. Impaired relaxation is associated with prolonged filling intervals whereas restrictive filling is characterized by increased filling rates.

Project description:BackgroundLeft ventricular (LV) contractility and compliance are derived from pressure-volume (PV) loops during dynamic preload reduction, but reliable simultaneous measurements of pressure and volume are challenging with current technologies. We have developed a method to quantify contractility and compliance from PV loops during a dynamic preload reduction using simultaneous measurements of volume from real-time cardiovascular magnetic resonance (CMR) and invasive LV pressures with CMR-specific signal conditioning.MethodsDynamic PV loops were derived in 16 swine (n = 7 naïve, n = 6 with aortic banding to increase afterload, n = 3 with ischemic cardiomyopathy) while occluding the inferior vena cava (IVC). Occlusion was performed simultaneously with the acquisition of dynamic LV volume from long-axis real-time CMR at 0.55 T, and recordings of invasive LV and aortic pressures, electrocardiogram, and CMR gradient waveforms. PV loops were derived by synchronizing pressure and volume measurements. Linear regression of end-systolic- and end-diastolic- pressure-volume relationships enabled calculation of contractility. PV loops measurements in the CMR environment were compared to conductance PV loop catheter measurements in 5 animals. Long-axis 2D LV volumes were validated with short-axis-stack images.ResultsSimultaneous PV acquisition during IVC-occlusion was feasible. The cardiomyopathy model measured lower contractility (0.2 ± 0.1 mmHg/ml vs 0.6 ± 0.2 mmHg/ml) and increased compliance (12.0 ± 2.1 ml/mmHg vs 4.9 ± 1.1 ml/mmHg) compared to naïve animals. The pressure gradient across the aortic band was not clinically significant (10 ± 6 mmHg). Correspondingly, no differences were found between the naïve and banded pigs. Long-axis and short-axis LV volumes agreed well (difference 8.2 ± 14.5 ml at end-diastole, -2.8 ± 6.5 ml at end-systole). Agreement in contractility and compliance derived from conductance PV loop catheters and in the CMR environment was modest (intraclass correlation coefficient 0.56 and 0.44, respectively).ConclusionsDynamic PV loops during a real-time CMR-guided preload reduction can be used to derive quantitative metrics of contractility and compliance, and provided more reliable volumetric measurements than conductance PV loop catheters.

Project description:BACKGROUND:Investigation of the myocardial strain characteristics of the left ventricular non-compaction (LVNC) phenotype with cardiovascular magnetic resonance (CMR) feature tracking. METHODS:CMR cine balanced steady-state free precession data sets of 59 retrospectively identified LVNC phenotype patients (40?years, IQR: 28-50?years; 51% male) and 36 healthy subjects (39?years, IQR: 30-47?years; 44% male) were evaluated for LV volumes, systolic function and mass. Hypertrabeculation in patients and healthy subjects was evaluated against established CMR diagnostic criteria. Global circumferential strain (GCS), global radial strain (GRS) and global longitudinal strain (GLS) were evaluated with feature-tracking software. Subgroup analyses were performed in patients (n =?25) and healthy subjects (n =?34) with normal LV volumetrics, and with healthy subjects (n =?18) meeting at least one LVNC diagnostic criteria. RESULTS:All LVNC phenotype patients, as well as a significant proportion of healthy subjects, met morphology-based CMR diagnostic criteria: non-compacted (NC): compacted myocardial diameter ratio?>?2.3 (100% vs. 19.4%), NC mass?>?20% (100% vs. 44.4%) and?>?25% (100% vs. 13.9%), and NC mass indexed to body surface area?>?15?g/m2 (100% vs. 41.7%). LVNC phenotype patients demonstrated reduced GRS (26.4% vs. 37.1%; p <?0.001), GCS (-?16.5% vs. -20.5%; p <?0.001) and GLS (-?14.6% vs. -17.1%; p <?0.001) compared to healthy subjects, with statistically significant differences persisting on subgroup comparisons of LVNC phenotype patients with healthy subjects meeting diagnostic criteria. GCS also demonstrated independent and incremental diagnostic value beyond each of the morphology-based CMR diagnostic criteria. CONCLUSIONS:LVNC phenotype patients demonstrate impaired strain by CMR feature tracking, also present on comparison of subjects with normal LV volumetrics meeting diagnostic criteria. The high proportion of healthy subjects meeting morphology-based CMR diagnostic criteria emphasizes the important potential complementary diagnostic value of strain in differentiating LVNC from physiologic hypertrabeculation.