Deep Learning Model to Quantify Left Atrium Volume on Routine Non-Contrast Chest Ct and Predict Adverse Outcomes

Abstract

Background: Non-contrast chest CTs (NCCT) are performed routinely for coronary artery calcium (CAC) scoring and lung cancer screening. However, a large amount of noncoronary and nonpulmonary data from these scans remain unassessed. We aimed to validate a deep learning model to automatically segment and measure left atrial (LA) volumes from routine NCCT and evaluate prediction for cardiovascular outcomes. Methods: We retrospectively evaluated 273 patients (median age 69 years, 55.5% male) who underwent a routine non-ECG gated NCCT for lung cancer screening. LA volumes were quantified by three expert cardiothoracic radiologists and a prototype AI algorithm. LA volumes were then indexed to the body surface area (BSA). Expert and AI LA volume index (LAVi) were compared and used to predict cardiovascular outcomes within five years. Logistic regression with appropriate univariate statistics were used for modelling outcomes. Findings: There was excellent correlation between AI and expert results with an LAV intraclass correlation of 0.950 (0.936-0.960). Bland-Altman plot demonstrated the AI underestimated LAVi by a mean 5.86 mL/m 2 . AI-LAVi was associated with new-onset atrial fibrillation (AUC 0.86; OR 1.12, 95% CI 1.08-1.18, p<0.001), HF hospitalization (AUC 0.90; OR 1.07, 95% CI 1.04-1.13, p<0.001), and MACE (AUC 0.68; OR 1.04, 95% CI 1.01-1.07, p=0.01). Interpretation: This novel deep learning algorithm for automated measurement of LA volume on routine NCCT had excellent agreement with manual quantification. AI-LAVi is significantly associated with new-onset atrial fibrillation, HF hospitalization, and MACE within 5 years. Funding: None to declare. Declaration of Interest: AJ, MAG, PS, and PS are employees of Siemens Healthineers. UJS has received institutional research support and/or honoraria for speaking and consulting from Astellas, Bayer, Bracco, Elucid BioImaging, General Electric, Guerbet, HeartFlow Inc., and Siemens Healthineers. TE has received a speaker fee and travel support from Siemens Healthineers. MEF reports research support from Boston Scientific, Biosense Webster, and Medtronic. JRB owns equity at YellowDot Innovations, LLC., and has received institutional research support and/or honoraria for consulting from Siemens Healthineers. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.