Diagnostic Performance of Machine Learning for Systemic Lupus Erythematosus: Systematic Review and Meta-Analysis.
Bingduo Wang, Zichao Wang, Yang Liu et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the diagnostic performance of machine learning (ML) and deep learning (DL) models for systemic lupus erythematosus (SLE) classification, lupus nephritis (LN) diagnosis, and neuropsychiatric SLE discrimination. The analysis included 29 studies, revealing a pooled sensitivity of 0.91 and specificity of 0.94 for SLE classification. The study highlights the need for improved validation and methodological rigor in future research.
Key findings
- Pooled sensitivity for SLE classification was 0.91 (95% CI 0.86-0.94).
- Pooled specificity for SLE classification was 0.94 (95% CI 0.91-0.96).
- Deep learning models demonstrated a sensitivity of 0.93 and specificity of 0.95.
- Only 31% of studies performed independent external validation.
Why it matters
Accurate and early diagnosis of systemic lupus erythematosus is crucial for effective management. Understanding the performance of ML and DL models can inform future diagnostic strategies, although current limitations must be addressed.
What to keep in mind
All included studies were retrospective, and 75.9% had high or unclear risk of bias.