Machine learning-based prognostic prediction for acute ischemic stroke using whole-brain and infarct multi-PLD ASL radiomics
Abstract Introduction Accurate early prognostic prediction for acute ischemic stroke (AIS) is essential for guiding personalized treatment. This study aimed to assess the predictive value of radiomics features from whole-brain and infarct cerebral blood flow (CBF) images using multiple post-labeling...
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2025-07-01
|
| Series: | BMC Medical Imaging |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12880-025-01807-w |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Introduction Accurate early prognostic prediction for acute ischemic stroke (AIS) is essential for guiding personalized treatment. This study aimed to assess the predictive value of radiomics features from whole-brain and infarct cerebral blood flow (CBF) images using multiple post-labeling delay arterial spin labeling (multi-PLD ASL), and to develop a prediction model incorporating clinical risk factors. Methods Radiomics features were extracted from the whole-brain and infarct regions based on multi-PLD ASL CBF images of 110 AIS patients. Five machine learning algorithms were used to construct radiomics models (whole-brain, infarct, and combined), clinical models, and comprehensive models integrating radiomics and clinical data. Model performance and clinical utility were assessed using receiver operating characteristic and decision curve analyses. Model stability was evaluated via 5000 permutation tests, and differences in the area under the curve (AUC) were compared using the DeLong test. Shapley Additive exPlanation was used to interpret feature contributions. Results The whole-brain and infarct radiomics models showed similar predictive performance. The combined radiomics models generally outperformed the infarct-only models. Additionally, no significant differences were observed between the combined radiomics models and the clinical models across the five algorithms. The comprehensive models, which integrated both radiomics and clinical features, demonstrated superior performance compared to both the clinical and combined radiomics models. Among all models, the comprehensive model based on a support vector machine achieved the highest predictive performance (AUC = 0.904). Its predictive capability was primarily driven by the baseline National Institutes of Health Stroke Scale score, age, infarct shape features, and higher-order statistical and texture features derived from both infarct and whole-brain CBF images. Conclusions The whole-brain CBF radiomics features of multi-PLD ASL can replace the infarct radiomics features that need to be delineated. Combined radiomics models outperformed infarct-only models and performed similarly to clinical models, which is applicable to AIS patients with incomplete clinical data. The comprehensive model integrates multi-PLD ASL CBF radiomics and clinical data, providing a safe and accurate tool for early prognosis prediction in AIS patients. |
|---|---|
| ISSN: | 1471-2342 |