The role of artificial intelligence in the diagnosis and prognosis of traumatic brain injury based on brain CT scans: a systematic review.
Eleni Romeo, Petros Baxevanidis, George A Alexiou et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review evaluated the role of artificial intelligence (AI) in diagnosing and predicting outcomes of traumatic brain injury (TBI) using brain CT scans. It included 22 studies, revealing that while diagnostic performance was generally high with AUC values approaching or exceeding 0.90, prognostic performance varied and was often biased due to lack of external validation.
Key findings
- Diagnostic performance was generally high, with many studies reporting AUC values approaching or exceeding 0.90.
- Prognostic performance showed moderate to high discrimination but substantial heterogeneity across studies.
- Only 9 studies incorporated independent external validation, raising concerns about generalizability.
- Most studies were at high risk of bias, particularly in patient selection.
Why it matters
Understanding the capabilities and limitations of AI in TBI diagnosis and prognosis is crucial for clinicians, as it may enhance decision-making and patient outcomes. However, caution is warranted due to potential biases and the need for further validation.
What to keep in mind
Most studies relied on retrospective designs and lacked independent external validation, limiting generalizability.