Sex-specific machine learning improves prediction of incident and prevalent COPD.
Sex-specific AI models predict COPD risk better than one-size-fits-all approaches, particularly improving detection in underdiagnosed women.
Using the CanCOLD cohort and external SPIROMICS validation, sex-specific CT-based machine learning models substantially outperformed combined-sex models in predicting both prevalent and incident COPD, particularly for women, by selecting parenchymal texture and lung-shape features that combined-sex models do not prioritize. These findings argue for adopting sex-disaggregated AI prediction strategies in COPD screening to improve early detection and risk stratification in females who are historically underdiagnosed.
What the study was
- Study design
- Retrospective cohort with external validation
- Population
- 1283 CanCOLD participants (internal) + 1840 SPIROMICS participants (external validation); CT imaging + spirometry at baseline and follow-up
- Sample size
- 3123
- Category
- Diagnostics
- Maturity
- Validated
- Journal
- Chest
Why it surfaced
Externally validated sex-specific ML models for COPD prediction with clinically significant AUC improvements; highlights actionable sex-disaggregation opportunity in current diagnostic AI pipelines.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.