Quantitative estimation of pulmonary artery wedge pressure from chest radiographs by a regression convolutional neural network

Introduction

Recent studies reported that a convolutional neural network (CNN; a deep learning model) can detect elevated pulmonary artery wedge pressure (PAWP) from chest radiographs, the diagnostic images most commonly used for assessing pulmonary congestion in heart failure. However, no method has been published for quantitatively estimating PAWP from such radiographs. We hypothesized that a regression CNN, an alternative type of deep learning, could be a useful tool for quantitatively estimating PAWP from chest radiographs in cardiovascular diseases.

Methods

We retrospectively enrolled 936 patients with cardiovascular diseases who had undergone right heart catheterization (RHC) and chest radiography and estimated PAWP by constructing a regression CNN based on the VGG16 model. We randomly categorized 80% of the data as training data (training group, n=748) and 20% as test data (test group, n=188). Moreover, we tuned the learning rate–one of the model parameters–by 5-hold cross-validation of the training group. Correlations between PAWP measured by RHC (ground truth [GT] PAWP) and PAWP derived from the regression CNN (estimated PAWP) were tested. To visualize how the regression CNN assessed the images, we created a regression activation map (RAM), a visualization technique for regression CNN.

Results

Estimated PAWP correlated significantly with GT PAWP in both the training (r=0.76, P<0.001) and test group (r=0.62, P<0.001). Bland-Altman plots found a mean (SEM) difference between GT and estimated PAWP of −0.23 (0.16) mm Hg in the training and −0.05 (0.41) mm Hg in the test group. The RAM showed that our regression CNN model estimated high PAWP by focusing on the cardiomegaly and pulmonary congestion. In the test group, the area under the curve (AUC) for detecting elevated PAWP (≥18 mm Hg) produced by the regression CNN model was similar to the AUC of an experienced cardiologist (0.86 vs 0.83, respectively; P=0.24).

Conclusion

This proof-of-concept study shows that regression CNN can quantitatively estimate PAWP from standard chest radiographs in cardiovascular diseases.

Funding Acknowledgement

Type of funding sources: Private company. Main funding source(s): The Bayer Academic Support

P1828Aortic calcification detected by computed tomography and aortic vulnerable plaques: aortic angioscopy study

Background

Aortic calcification is associated with atherosclerotic risk factors and an increased risk of death and cardiovascular disease. However, the relationships aortic calcification and aortic plaque instability are not yet elucidated. Recently, some reports showed non-obstructive aortic angioscopy seemed to visualize atherosclerotic changes of aortic wall more clearly compared with computed tomography (CT). The purpose of this study was to evaluate whether aortic calcification is associated with aortic vulnerable plaques in patients with cardiovascular disease.

Methods

We investigated 60 consecutive patients with confirmed or suspected coronary artery disease who underwent both aortic angioscopy and CT. The AC volume (ACV) was measured using the volume-rendering method by extracting the area >130 HU within the whole aorta. ACV index (ACVI) was defined as ACV divided by the body surface area. We evaluated the number of ruptured plaque (RP), ulceration and fissure by aortic angioscopy in the whole aorta. We excluded 4 hemodialysis patients. All patients were divided into the median value of ACVI.

Results

The mean age of patients was 68±10. The median of ACVI was 10.7 ml/m2 [3.9–22.7]. High ACVI patients had significantly greater number of RP, ulceration and atheromatous plaques detected by aortic angioscopy compared with those of low ACVI (2.2±2.7 vs 0.8±1.1, p=0.033, 1.6±1.2 vs 0.9±1.0, p=0.041, 4.0±3.1 vs 1.9±1.8, p=0.009, respectively). Furthermore, the patients without aortic calcification did not have RP at all. In a multivariate model, the number of the atheromatous plaques was independently associated with high ACVI (odds ratio 1.57, 95% confidence interval 1.07–2.69, p=0.018)

Conclusions

Aortic calcification detected by CT was related to aortic vulnerable plaques in patients with cardiovascular disease.

6125High wall shear stress predicts plaque rupture of the aortic arch: computational fluid dynamics model and non-obstructive general angioscopy study

Background

Wall shear stress (WSS) has been considered as a major determinant of aortic atherosclerosis. Recently, non-obstructive general angioscopy (NOGA) was developed to be able to visualize a variety of its atherosclerotic pathology, including in vivo ruptured plaque (RP) in the aorta. We, therefore, investigated the relationship between NOGA derived RP in the aortic arch and the stereographic distribution of WSS by using computational fluid dynamics modeling (CFD) on three-dimensional CT angiography (3D-CT).

Methods

We investigated 30 consecutive patients who underwent 3D-CT before and NOGA during coronary angiography. WSS in the aortic arch was measured with an application of CFD based on finite element method by using uniform inlet and outlet flow conditions. Aortic RP was detected by NOGA.

Results

The maximum and mean values of WSS were 67.2±29.2 Pa and 2.4±0.6 Pa. A total of 18 RPs was detected by NOGA. The patients with a distinct RP showed a significantly higher maximum WSS in the whole aortic arch, and the greater and lesser curvature of the aortic arch than those without it (73.3±29.0 Pa vs 50.4±15.2 Pa, p=0.035, 95.0±27.5 Pa vs 42.8±25.2 Pa, p=0.003, 70.8±29.3 Pa vs 46.1±11.9 Pa, p=0.013, respectively), whereas there was no significant difference in the mean WSS between those with and without it. In a multivariate analysis, the maximum value of WSS was an independent predictor of RP in the aortic arch (odds ratio 1.05, 95% confidence interval 1.01–1.13, p=0.019).

Representative picture of WSS and NOGA

Conclusions

Aortic RP detected by NOGA was strongly associated with the higher maximum WSS in the aortic arch derived by CFD using 3D-CT. Maximum WSS may explain the underlying mechanism of not only aortic atherosclerosis, but also aortic RP.