Key result
CCTA deep learning model outperforms cardiologists in predicting PCI atherectomy use with ~0.79 AUC.
Why the study?
Severe coronary calcification complicates PCI and often necessitates atherectomy-based debulking devices that increase procedural complexity and risk. This study sought to develop and externally validate a Transformer-based deep learning model using preprocedural CCTA to predict atherectomy device use and compare its performance with blinded human readers.
Does a Transformer-based deep learning model using preprocedural CCTA accurately predict atherectomy device use during PCI compared to blinded human readers in patients with calcified coronary lesions?
Cohort (n=419)
Blinded human readers
Yes
Does a Transformer-based deep learning model using preprocedural CCTA accurately predict atherectomy device use during PCI compared to blinded human readers in patients with calcified coronary lesions?
A Transformer-based deep learning model using preprocedural CCTA can predict the need for atherectomy devices during PCI with higher accuracy than general cardiologists and comparable accuracy to PCI experts, potentially aiding preprocedural planning.
Answers are grounded in this paper and cite their sources.
May aid preprocedural PCI planning for calcified lesions; leaves open prospective validation before clinical adoption.
Aims Severe coronary calcification complicates percutaneous coronary intervention (PCI), often necessitating atherectomy-based debulking devices that increase procedural complexity and risk. We aimed to develop and externally validate a Transformer-based deep learning model using preprocedural cardiac computed tomography angiography (CCTA) to predict atherectomy device use during PCI and compare its performance with that of blinded human readers, including general cardiologists and PCI experts. Methods and results This retrospective multicentre study included 498 calcified coronary lesions from 419 patients who underwent PCI after preprocedural CCTA. Data from 12 institutions were used for training and validation with four-fold cross-validation, and data from two independent institutions (97 lesions) served as an external test cohort. The model combined a convolutional neural network feature extractor with Transformer-based multiple instance learning, and clinical variables were evaluated for incremental value. In the external test cohort, the CCTA-only model achieved an area under the receiver operating characteristic curve (AUC) of 0.788 (95% confidence interval [CI], 0.691–0.873) for predicting atherectomy device use. Adding clinical variables resulted in an AUC of 0.805 (95% CI, 0.713–0.886), indicating limited incremental value. In the reader comparison, the model showed higher accuracy than general cardiologists (0.742 vs. 0.619; P = 0.029) and numerically similar accuracy to PCI experts (0.742 vs. 0.670; P = 0.230). Conclusion A Transformer-based deep learning model using preprocedural CCTA showed clinically relevant performance for predicting atherectomy device use during PCI in an independent external test cohort. These findings suggest that the model may support lesion assessment and procedural planning.
No takes yet. Share an insight, caveat, or question.
Hino et al. (2026) conducted a cohort in Severe coronary calcification undergoing percutaneous coronary intervention (PCI) (n=419). Transformer-based deep learning model using preprocedural CCTA vs. Blinded human readers (general cardiologists and PCI experts) was evaluated on Prediction of atherectomy device use during PCI (AUC 0.788, 95% CI 0.691-0.873, p=0.029). A CCTA-based Transformer deep learning model predicted atherectomy use during PCI with an AUC of 0.788 (95% CI 0.691-0.873), outperforming general cardiologists (accuracy 0.742 vs 0.619; P=0.029).
Synapse has enriched 10 related papers. Consider them for comparative context: