Bisphosphonates Reduce Fracture Risk in Prostate Cancer Patients on Androgen Deprivation Therapy, suggests study

Training was performed using a multitask deep learning framework that could process both diagnostic classification and echocardiographic parameter estimation concurrently.

The performance of PanEcho was to be assessed using two main measures:

• Area Under the Receiver Operating Characteristic Curve (AUC): For diagnostic condition classification

• Mean Absolute Error (MAE): For echocardiographic parameter estimation

These AI-derived results were compared to cardiologists' interpretations obtained from standard clinical practice.

Key Findings

Diagnostic Accuracy:

• PanEcho performed 18 tasks of diagnostic classification with a median AUC of 0.91 (interquartile range [IQR]: 0.88–0.93) in internal validation.

Parameter Estimation:

• For 21 measurements of echocardiography, the median normalized mean absolute error was 0.13 (IQR: 0.10–0.18).

• Left ventricular ejection fraction (LVEF): Estimated with mean absolute error of 4.2% (internal) and 4.5% (external).

• Moderate or greater left ventricular systolic dysfunction detection: AUC of 0.98 (internal) and 0.99 (external).

• Right ventricular systolic dysfunction: AUC of 0.93 (internal) and 0.94 (external).

• Severe aortic stenosis: AUC of 0.98 (internal) and 1.00 (external).

Performance in Limited Protocols:

• Abbreviated TTE cohort: 15 diagnosis tasks, median AUC of 0.91 (IQR: 0.87–0.94)

• Emergency department point-of-care cohort: 14 diagnosis tasks, median AUC of 0.85 (IQR: 0.77–0.87).

The study showed the reliability of the AI model PanEcho in reading broad echocardiographic parameters and diagnostic labels for high accuracy with diverse clinical settings. These findings establish the feasibility of integrating PanEcho into standard cardiovascular imaging workflow particularly as an assistant to help streamline and standardize echocardiogram reporting.

Reference:

Jones, C., Dutey-Magni, P., Murphy, L. R., Murray, M. L., Brown, J. E., McCloskey, E., Brown, M., Amos, C. L., Gilbert, D. C., Jones, R. J., Cross, W., Matheson, D., Millman, R., Parmar, M. K. B., Attard, G., Sydes, M. R., Brown, L. C., James, N. D., Clarke, N. W., Sachdeva, A., … STAMPEDE Trial Investigators (2025). Fracture-related hospitalisations in newly diagnosed high-risk localised or metastatic hormone-sensitive prostate cancer: secondary analysis of the STAMPEDE phase 3 trials of docetaxel and zoledronic acid using healthcare systems data. Annals of oncology : official journal of the European Society for Medical Oncology, S0923-7534(25)00854-3. Advance online publication. https://doi.org/10.1016/j.annonc.2025.07.005