Automatic thoracic aorta calcium quantification using deep learning in non-contrast ECG-gated CT images

Thoracic aorta calcium (TAC) can be assessed from cardiac computed tomography (CT) studies to improve cardiovascular risk prediction. The aim of this study was to develop a fully automatic system to detect TAC and to evaluate its performance for classifying the patients into four TAC risk categories. The method started by segmenting the thoracic aorta, combining three UNets trained with axial, sagittal and coronal CT images. Afterwards, the surrounding lesion candidates were classified using three combined convolutional neural networks (CNNs) trained with orthogonal patches. Image datasets included 1190 non-enhanced ECG-gated cardiac CT studies from a cohort of cardiovascular patients (age 57 ± 9 years, 80% men, 65% TAC > 0). In the test set (N = 119), the combination of UNets was able to successfully segment the thoracic aorta with a mean volume difference of 0.3 ± 11.7 ml (<6%) and a median Dice coefficient of 0.947. The combined CNNs accurately classified the lesion candidates and 87% of the patients (N = 104) were accurately placed in their corresponding risk categories (Kappa = 0.826, ICC = 0.9915). TAC measurement can be estimated automatically from cardiac CT images using UNets to isolate the thoracic aorta and CNNs to classify calcified lesions.

Echocardiographic measurements of the left heart chambers size in a large cohort of subjects: comparison of body surface area and height indexation to account for effects of obesity

BACKGROUND

The assessment of cardiac chamber size in the obese population is a challenging subject. Values usually indexed to body surface area (BSA) are smaller in obese subjects and prone to overcorrection. This study aimed to find reference thresholds to account for effects of obesity from a large cohort of patients and to evaluate the indexation by height as an alternative to BSA.

METHODS

The last 10-years records of our Echocardiography Unit were retrospectively analyzed and 14,007 subjects without known cardiac disease were included (45±15 years, 54% women, 20% obese). Measurements included left atrial diameter, area and volume (LAD, LAA and LAV), left ventricular (LV) end-diastolic (LVEDD) and end-systolic diameters (LVESD), aortic root diameter (ARD) and LV mass (LVM). Absolute, BSA and height indexed maximum thresholds (mean+1.96.SD) were calculated. An allometric indexation of the form variable/Height^β was tested. Correlation coefficients between indexed and absolute values were calculated to evaluate their proportional association (ideally r=1). Correlations between indexed values and body size represented residual associations to be minimized (ideally r=0).

RESULTS

The strongest association of echocardiographic measurements with body size was observed for BSA (r=0.36-0.63), whereas the isometric and allometric height models showed lower comparable values (r=0.28-0.48). Positive correlations with BMI were mostly observed for LA size (r≈0.36) and LVM (r≈0.36) measurements. Scaling exponents β for the allometric height indexation were 1.72 for LAV and 2.33 for LVM. Correlations between indexed and absolute values were higher for height than BSA (0.80-0.98 vs 0.44-0.92). Correlations between indexed values and height were closer to zero than for BSA, particularly using the allometric model. The overcorrection observed with increasing obesity class after BSA indexation was avoided after height indexation.

CONCLUSIONS

Unlike BSA, height indexing provided an adequate body size scaling of left heart chambers size avoiding overcorrection using allometric models in particular.

Thoracic Aorta Calcium Detection and Quantification Using Convolutional Neural Networks in a Large Cohort of Intermediate-Risk Patients

Arterial calcification is an independent predictor of cardiovascular disease (CVD) events whereas thoracic aorta calcium (TAC) detection might anticipate extracoronary outcomes. In this work, we trained six convolutional neural networks (CNNs) to detect aortic calcifications and to automate the TAC score assessment in intermediate CVD risk patients. Cardiac computed tomography images from 1415 patients were analyzed together with their aortic geometry previously assessed. Orthogonal patches centered in each aortic candidate lesion were reconstructed and a dataset with 19,790 images (61% positives) was built. Three single-input 2D CNNs were trained using axial, coronal and sagittal patches together with two multi-input 2.5D CNNs combining the orthogonal patches and identifying their best regional combination (BRC) in terms of lesion location. Aortic calcifications were concentrated in the descending (66%) and aortic arch (26%) portions. The BRC of axial patches to detect ascending or aortic arch lesions and sagittal images for the descending portion had the best performance: 0.954 F1-Score, 98.4% sensitivity, 87% of the subjects correctly classified in their TAC category and an average false positive TAC score per patient of 30. A CNN that combined axial and sagittal patches depending on the candidate aortic location ensured an accurate TAC score prediction.

4D flow MRI: impact of region of interest size, angulation and spatial resolution on aortic flow assessment

Objectives.In cardiovascular magnetic resonance, the 3D time-resolved phase-contrast technique, also known as 4D flow, is gaining increasing attention due to applications that exploit three-directional velocity encoding throughout the cardiac cycle. Blood flow volume assessment usually requires an expert to draw regions of interest (ROI) around the vessel cross section, whereas the errors involved in this estimation have not been thoroughly investigated. Our objective is to quantify the influence of ROI sizing, angulation and spatial resolution of the reconstructed plane employed in blood flow measurements using 4D flow.Approach.Three circular ROIs were drawn around the ascending, arch and descending aorta of healthy volunteers (n= 27) and patients with a dilated ascending aorta or bicuspid valve (n= 37). We applied systematic changes of ROI diameter (up to ±10%), tilt angle (up to ±25°) and spatial resolution (from 0.25 to 2 mm) of the reconstructed oblique planes, calculating the effects on net, forward and backward blood flow volumes.Main results.Patients had a larger ascending aorta than healthy volunteers with similar ages and male sex proportion (60 ± 15 y.o. vs 58 ± 16 y.o. and 84% vs 70%, respectively). Higher forward and backward flow volumes were observed in the ascending aorta and the aortic arch of the patients with respect to controls (p< 0.001), whereas net volumes were similar: 74.0 ± 20.8 ml versus 75.7 ± 21.8 ml (p= 0.37), respectively. The ascending aorta was the most sensitive to ROI modifications. Changes of ±10% in the ROI diameter and ±25° in tilt angles produced flow volume differences of up to 9 ml (10%) and 18 ml (15%) in controls and patients, respectively. Modifying the reconstructed planes spatial resolution produced flow volume changes below 2 ml.Significance.Since the setting of the ROI size and plane angle could produce errors that represent up to 20% of the forward and/or backward aortic flow volume, a good standardization for vessel segmentation and plane positioning is desirable.