Predicting immune reconstitution after antiretroviral therapy in HIV/AIDS using ensemble machine learning: a real-world study
Machine learning predicts how quickly HIV patients rebuild immune cells after starting treatment without needing frequent CD4 testing.
A stacking ensemble of four gradient tree algorithms predicts longitudinal CD4+ T-cell reconstitution (R²=0.768) in 5,436 ART-treated HIV patients without relying on baseline immunological measures, substantially outperforming a Robust Transformer baseline. This framework enables data-driven personalized monitoring for HIV patients in resource-variable settings where CD4 testing may be irregular.
What the study was
- Study design
- Retrospective cohort with heterogeneous stacking ensemble ML (XGBoost, LightGBM, RF, GB + Ridge meta-learner)
- Population
- People living with HIV on ART (n=5,436 training, n=1,088 test), Xi'an, China, 2016–2025
- Sample size
- 6524
- Category
- Diagnostics
- Maturity
- Exploratory
- Journal
- Frontiers in Immunology
Why it surfaced
Large real-world ML cohort for HIV immune monitoring; unsolicited find with moderate relevance to AI/ML in clinical diagnostics watchlist adjacency.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.