Accuracy of Deep Learning in Detecting Cerebral Microbleeds: Systematic Review and Meta-Analysis.
J Med Internet ResSep 21, 2026 (epub)
Yue Feng, Lei Zheng, Baiwen Zhang et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the diagnostic accuracy of deep learning (DL) models for detecting cerebral microbleeds (CMBs) using MRI in five studies. The findings indicated high sensitivity and specificity, with patient-level sensitivity at 89% and specificity at 86%, while lesion-level sensitivity reached 96% and specificity 98%.
Key findings
- At the patient level, sensitivity was 0.89 and specificity was 0.86.
- At the lesion level, sensitivity was 0.96 and specificity was 0.98.
- In the direct extraction subgroup, sensitivity was 0.98 and specificity was 0.98.
- The diagnostic odds ratio at the lesion level was 1061.
Why it matters
The use of deep learning models for detecting cerebral microbleeds could enhance diagnostic accuracy and efficiency, potentially reducing reliance on expert interpretation and associated costs in clinical practice.