Cardio-oncology & otherDiagnostic or prognostic study
Machine Learning Model Using Pre-Cancer Therapy Cardiac Magnetic Resonance Images to Predict Cancer Therapy-Related Cardiac Dysfunction
- Diagnostic accuracy study
- pooled from 3 prospective studies
- internal and external validation
Authors' conclusion
A DL model using pre-cancer therapy CMR short-axis cines can predict future CTRCD risk better than clinical or manually quantified imaging models.
Study context
- Population
- Women with HER2+ breast cancer receiving anthracyclines and trastuzumab (Toronto and Calgary, Canada)176 internal, 53 external
- Participants
- 229 women
- Comparison
- Pre-cancer therapy cardiac magnetic resonance (CMR) deep learning model vs HFA-ICOS risk score, CMR and echocardiographic models
- Outcome
- Prediction of cancer therapy-related cardiac dysfunction (CTRCD) (defined by CMR)
Performance
As reported in the abstract
Prediction of CTRCD
| Test or model | AUC | F1 score |
|---|---|---|
| Deep learning (DL) model | 0.85 (95% CI: 0.69-0.97) | 0.69 (95% CI: 0.47-0.86) |
| HFA-ICOS | 0.66 (95% CI: 0.53-0.75) | 0.56 (95% CI: 0.47-0.61) |
| CMR | 0.59 (95% CI: 0.41-0.73) | 0.20 (95% CI: 0.00-0.42) |
| Echocardiography | 0.62 (95% CI: 0.44-0.78) | 0.36 (95% CI: 0.12-0.62) |
Further findings
| Outcome | DL model | HFA-ICOS | CMR | Echocardiography |
|---|---|---|---|---|
| Prediction of CTRCD, external validation (Calgary) | ||||
| AUC | 0.80 (95% CI: 0.66-0.92) | Not reported in the abstract | Not reported in the abstract | Not reported in the abstract |
| F1 score | 0.55 (95% CI: 0.24-0.77) | Not reported in the abstract | Not reported in the abstract | Not reported in the abstract |
Original publication
Machine Learning Model Using Pre-Cancer Therapy Cardiac Magnetic Resonance Images to Predict Cancer Therapy-Related Cardiac Dysfunction. JACC Cardiovasc Imaging. Published 28 May 2026. doi:10.1016/j.jcmg.2026.04.007
Summarised from the published abstract; every number is checked against it