Deep convolutional neural networks to predict cardiovascular risk from computed tomography

Abstract

Coronary artery calcium is an accurate predictor of cardiovascular events. While it is visible on all computed tomography (CT) scans of the chest, this information is not routinely quantified as it requires expertise, time, and specialized equipment. Here, we show a robust and time-efficient deep learning system to automatically quantify coronary calcium on routine cardiac-gated and non-gated CT. As we evaluate in 20,084 individuals from distinct asymptomatic (Framingham Heart Study, NLST) and stable and acute chest pain (PROMISE, ROMICAT-II) cohorts, the automated score is a strong predictor of cardiovascular events, independent of risk factors (multivariable-adjusted hazard ratios up to 4.3), shows high correlation with manual quantification, and robust test-retest reliability. Our results demonstrate the clinical value of a deep learning system for the automated prediction of cardiovascular events. Implementation into clinical practice would address the unmet need of automating proven imaging biomarkers to guide management and improve population health.

Automated coronary calcium scoring using deep learning with multicenter external validation

“…Our model uses a single convolutional neural network (CNN) for an end-toend approach. Most significantly, all deep learning models 26,28,29,31,32 on CAC scoring using non-gated unenhanced chest CTs reported to date have used manual scoring on non-gated chest CTs solely as the reference standard, which may be inadequate. Most notably, a recent study by Zeleznik et al 32 demonstrated substantial agreement between automated and manual stratification of CAC scores into one of four risk buckets in a multi-center trial cohort comprised of over 20k asymptomatic individuals.…”

“…Most significantly, all deep learning models 26,28,29,31,32 on CAC scoring using non-gated unenhanced chest CTs reported to date have used manual scoring on non-gated chest CTs solely as the reference standard, which may be inadequate. Most notably, a recent study by Zeleznik et al 32 demonstrated substantial agreement between automated and manual stratification of CAC scores into one of four risk buckets in a multi-center trial cohort comprised of over 20k asymptomatic individuals. However, the comparison here was also made to manual quantitation performed on non-gated studies as opposed to the current standard of care to quantify CAC, a gated coronary CT exam.…”

“…Additionally, while some studies 26 – 32 have reported the feasibility of automated methods to quantify coronary calcium from non-gated unenhanced chest CTs, they are either not end-to end 27 , 28 or do not use deep learning methods 27 , 30 . Most importantly, for validation, none of these studies 26 – 32 curated a dataset comprised of a strong clinical reference standard for ground truth, which calls into question the clinical claims and require further validation.…”

“…Though many semi-automated and few automated calcium scoring methods using gated CTs have been proposed in the literature, to our knowledge, no study to date has reported fully automated vessel-specific coronary calcium scoring using an end-to-end deep learning architecture with prospective validation and multi-center external validation [23][24][25] . Additionally, while some studies [26][27][28][29][30][31][32] have reported the feasibility of automated methods to quantify coronary calcium from non-gated unenhanced chest CTs, they are either not end-to end 27,28 or do not use deep learning methods 27,30 . Most importantly, for validation, none of these studies [26][27][28][29][30][31][32] curated a dataset comprised of a strong clinical reference standard for ground truth, which calls into question the clinical claims and require further validation.…”

Position paper of the EACVI and EANM on artificial intelligence applications in multimodality cardiovascular imaging using SPECT/CT, PET/CT, and cardiac CT

“…van Velzen et al used DL to identify and quantify calcium on CT using 7240 participants, which included ECG-gated CT, diagnostic CT of the chest, PET attenuation correction CT, radiotherapy planning CT, and low-dose screening CT for lung cancer [ 51 ]. The resulting model had an intraclass correlation coefficient of 0.85–0.99 for the identification of CAC, leading to the prospect of routine automated quantification of calcification on thoracic CT. More recently, a study using 20,084 gated and non-gated cardiac CT scans developed a deep learning model to identify coronary calcification with excellent correlation with manual readers (r 0.92, p < 0.001) and test-retest stability (intra-class correlation 0.993, p < 0.001) [ 52 ].…”