Calcium scoring in low-dose ungated chest CT scans using convolutional long-short term memory networks

Abstract

We aimed to develop a novel deep-learning based method for automatic coronary artery calcium (CAC) quantification in low-dose ungated computed tomography attenuation correction maps (CTAC). In this study, we used convolutional long-short -term memory deep neural network (conv-LSTM) to automatically derive coronary artery calcium score (CAC) from both standard CAC scans and low-dose ungated scans (CT-attenuation correction maps). We trained convLSTM to segment CAC using 9543 scans. A U-Net model was trained as a reference method. Both models were validated in the OrCaCs dataset (n=32) and in the held-out cohort (n=507) without prior coronary interventions who had CTAC standard CAC scan acquired contemporarily. Cohen's kappa coefficients and concordance matrices were used to assess agreement in four CAC score categories (very low: <10, low:10-100; moderate:101-400 and high >400). The median time to derive results on a central processing unit (CPU) was significantly shorter for the conv-LSTM model- 6.18s (inter quartile range [IQR]: 5.99, 6.3) than for UNet (10.1s, IQR: 9.82, 15.9s, p<0.0001). The memory consumption during training was much lower for our model (13.11Gb) in comparison with UNet (22.31 Gb). Conv-LSTM performed comparably to UNet in terms of agreement with expert annotations, but with significantly shorter inference times and lower memory consumption.

Keywords: Convolutional LSTM; Coronary calcium scoring; PET CTAC; deep learning.

Figure 1.. Architecture of the convolutional long-short term deep neural network.

Our solution consists of two networks, first of which (A) was trained for segmentation of the heart silhouette while the second network (B) was trained to segment the coronary artery calcium using expert annotations. The heart mask was applied to the final CAC prediction to reduce spurious bone overcalling.

Figure 2.. Agreement between deep learning models predictions and expert readers in four calcium score classes for convLSTM and Unet.

First column – agreement between expert annotated CAC scans and DL scores obtained from the CAC scans. Second column – agreement between expert annotated CAC scans and DL prediction on contemporarily acquired CTAC maps CAC – coronary artery calcium, CTAC – CT attenuation correction, DL – deep learning

Figure 3.. Agreement between expert annotations and DL predictions on the OrCaCs dataset.

CAC – coronary artery calcium, CTAC – CT attenuation correction, DL – deep learning