Diagnostic performance and clinical applications of artificial intelligence for intracranial bleeding detection: A meta-analysis
Bibliographic record
Abstract
Introduction: Intracranial hemorrhage (ICH) is a neurological emergency with high mortality rates requiring timely diagnosis. While computed tomography (CT) remains the gold standard, diagnostic accuracy varies with radiologist experience and workload. This systematic review and meta-analysis aims to evaluate the diagnostic performance of AI algorithms in detecting ICH on CT imaging and to explore key considerations for their clinical implementation in emergency and teleradiology settings. Methods: We conducted a systematic review and meta-analysis following PRISMA-DTA guidelines, searching seven databases up to May 2025. Studies evaluating AI diagnostic accuracy for ICH detection on non-contrast CT scans were included. Quality assessment used QUADAS-2 criteria. Pooled estimates were calculated using random-effects models, with subgroup analyses by algorithm architecture and ICH subtype. Results: A total of 45 studies met the inclusion criteria, comprising 29 research algorithm evaluations (n = 185,847 patients) and 16 studies of commercial AI system implementations (n = 94,523 patients). Research algorithms demonstrated a pooled sensitivity of 0.890 (95 % CI: 0.839-0.942) and specificity of 0.926 (95 % CI: 0.899-0.954). Commercial AI systems exhibited slightly superior performance, with sensitivity of 0.899 (95 % CI: 0.858-0.940) and specificity of 0.951 (95 % CI: 0.928-0.974). Diagnostic accuracy varied notably across ICH subtypes, with epidural hemorrhage presenting the greatest detection challenge (difficulty score: 0.251). Among algorithmic designs, convolutional recurrent neural networks (CNN-RNNs) demonstrated the highest diagnostic performance. In real-world clinical implementation, AI integration demonstrated substantial workflow improvements: door-to-treatment decision time reduced by 26 % (92 → 68 min), critical case notification time decreased by 57 % (75 → 32 min), and triage accuracy improved by 8 % (86 %→94 %), directly impacting patient care pathways. Despite a 7-8 % sensitivity reduction compared to benchmark settings, these clinical benefits were consistent across implementations. Conclusions: AI algorithms demonstrate strong diagnostic performance in detecting ICH, with commercial systems demonstrating superior specificity compared to research models. Despite notable performance gaps in detecting certain hemorrhage subtypes, particularly epidural hemorrhage, the clinical benefits of AI integration, including improved workflow efficiency and reduced time to treatment decisions, are substantial. Future research should prioritize prospective validation and the development of algorithms tailored to enhance detection across challenging ICH subtypes.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.026 | 0.067 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.014 | 0.055 |
| Bibliometrics | 0.006 | 0.006 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.005 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".