Publications
Publications
Peer-reviewed abstracts and papers from the APEC Vox QCA programme.
-
Accepted abstract: ESC Digital & AI Summit 2026
A new abstract from the APEC Vox QCA programme has been accepted for the second ESC Digital & AI Summit. Its title, authors and citation appear here once it is presented and published.
-
AI-driven coronary stenosis detection versus IVUS reference standard
APEC International Ltd, Research and Development, London, United Kingdom
European Heart Journal – Digital Health Volume 7, Supplement 1, ztaf143.037
A deep-learning framework that segments the coronary tree on an angiogram, extracts the centreline, computes the diameter along it and quantifies stenosis, benchmarked against intravascular ultrasound. The segmentation model is a hybrid of MedSAM and VM-UNet. Segmentation was evaluated on ARCADE and on a mixed dataset; the IVUS concordance analysis was ongoing at the time of the abstract.
- IoU, mixed dataset
- 0.63
- Sensitivity
- 0.98
- Specificity
- 0.99
- Stenosis TPR / PPV
- 0.59 / 0.59
Figures are the abstract's: segmentation IoU 0.6308 with sensitivity 0.9772 and specificity 0.9903 on the mixed dataset (0.6303 / 0.9832 / 0.9933 on ARCADE alone); stenosis detection TPR 0.5867 and PPV 0.5911.
Cite
M. Aryafar, F. T. Tahavori, M. A. Aryafar. AI-driven coronary stenosis detection versus IVUS reference standard. European Heart Journal – Digital Health 7(Suppl 1):ztaf143.037 (2026). https://doi.org/10.1093/ehjdh/ztaf143.037
The reference-standard agreement study — our measurements against commercial 3D-QCA, case by case — is underway and will be reported here when it completes.