Interpretable machine learning for prediction of clinical outcomes in acute ischemic stroke
Problem
Predicting the 3-month functional outcome of acute ischemic stroke patients is needed to establish suitable treatment plans, but is clinically difficult.
Why it matters
Stroke prognosis directly shapes treatment decisions affecting patient survival and recovery.
What the AI did
Developed and validated machine learning models predicting 3-month functional outcome (favorable vs unfavorable mRS), using SHAP to interpret feature contributions, on a multicenter registry with two external validation sets.
Results
Derivation set 3,687 patients; external validation sets of 250 and 110 patients; model performance metrics not stated in provided abstract (truncated).
Impact
External validation across datasets suggests generalizable, interpretable outcome prediction, per the abstract.
Record
| Venue | Frontiers in Neurology |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Joonwon Lee, Kang Min Park, Seong-Ho Park |
| Citations | 33 (OpenAlex, at import time) |
| References | 35 |
| Model or method | ML methods with SHAP; specific algorithms not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3389/fneur.2023.1234046 |
| Open access | gold |
| Keywords | acute ischemic stroke, machine learning, prognosis, SHAP, external validation |
Importance rating
significant — thousands to millions affected
Stroke prognosis shapes treatment plans affecting survival/recovery; multicenter validated model, but abstract gives no disease scale.