Diagnostic or prognostic study
Deep Learning of Coronary Calcium Scores From Positron Emission Tomographic (PET)/Computed Tomographic (CT) Attenuation Maps Accurately Predicts Adverse Cardiovascular Events
- Diagnostic accuracy study
- external validation cohort
- Deep-learning (DL) model trained on 9,543 expert-annotated CT scans
Authors' conclusion
DL CTAC scores predict cardiovascular risk similarly to standard CAC scores quantified manually by experienced operators from dedicated electrocardiographically gated CAC scans and can be obtained almost instantly, with no changes to PET/CT scanning protocol.
Study context
- Population
- Patients undergoing PET/CT imaging (external cohort)2,737 with paired gated coronary artery calcium (CAC) scans
- Participants
- 4,331 patients
- Comparison
- DL computed tomographic attenuation correction (CTAC) CAC scores vs standard ECG-gated CAC scores
- Outcome
- MACE risk stratification by CAC score category (0, 1-100, 101-400, >400)
- Follow-up
- 4.3 years
Performance
MACE
Net reclassification improvement
−0.02
95% CI −0.11 to 0.07
By group
Standard
85%
NPV of zero CAC
DL CTAC
83%
NPV of zero CAC
As reported in the abstract
- DL CTAC scores: stepwise increase in MACE risk across CAC categories (HR up to 3.2)
- P < 0.001
Further findings
Automatic DL scoring time
- Automatic DL scoring required <6 seconds per scan
Original publication
Deep Learning of Coronary Calcium Scores From PET/CT Attenuation Maps Accurately Predicts Adverse Cardiovascular Events. JACC Cardiovasc Imaging. Published 14 September 2022. doi:10.1016/j.jcmg.2022.06.006
Summarised from the published abstract; every number is checked against it