Article Figures & Data
- Figure 1
The flow chart of FR-model development. The process of FR-Model development consisted of five parts: (a) Data preprocessing: select variables based on expert opinion and availability, select patients based on the inclusion criteria, handle missing and error values, scale all individual features to have a unit norm, and integrate low-level semantic information variables into higher-level variables. (b) Feature selection, the study adopted a two-stage feature selection pipeline (TFSP), including two steps: selection of variables with p<0.05, and selection of the top 10 important features using feature importance ranking. (c) Model training, the model was developed with selected features based on a 10-fold nested cross-validation (CV) framework, and 5 baseline machine learning models were fitted during the nested CV, including RF, gcForest, SVM, XGBoost and KNN. (d) Performance evaluation, the AUC was used to compare the performance of different models. Moreover, sensitivity, specificity, accuracy and precision were also considered as auxiliary indicators for the general evaluation of the forecasting model characteristics. According to the model predictive probability, two cut-off thresholds were adopted for the RESCUE-FR index to divide the patient population into the low-risk group, intermediate-risk group and high-risk group. (e) External Validation, the top 10 features with the highest frequency across all internal CV folds were selected as new model inputs and were used to retrain the model in the derivation cohort, whose performance was then evaluated in the validation cohort and compared with the proposed model in previous research. AUC, area under the curve; BMI, body mass index; FR, futile recanalisation; KNN, k-nearest neighbour; ML, machine learning; RF, random forest; ROC, receiver operating characteristics; SVM, support vector machines; GA, genetic algorithms; MICE, multiple imputation by chained equations; RESCUE-FR, Registration study for Critical Care of Acute Ischemic Stroke: Prediction of Futile Recanalisation.
- Figure 2
Performance of different models in RESCUE-RE using TFSP (10 features). According to 10 features selected with TFSP, 5 algorithms were used for each model. The performance of five algorithms in each model is shown in a–f. AUC, area under the curve; RESCUE-RE, Registration study for Critical Care of Acute Ischaemic Stroke after Recanalisation; TFSP, two-stage feature selection pipeline.
- Figure 3
Comparing the predictive power of the RESCUE-FR model with that of other models and risk scores in validation cohort. The performance of the proposed model in previous research was compared with the RESCUE-FR model in the validation cohort. The performance of the proposed model in the previous research was calculated according to the risk score formula provided by the research articles. AUC, area under the curve; RESCUE-FR, Registration study for Critical Care of Acute Ischemic Stroke: Prediction of Futile Recanalisation.
- Figure 4
Implementation of RESCUE-FR Index and the efficiency of risk stratification. (a) RESCUE-FR Index: An Artificial Intelligence Clinical Decision Tool for Forecasting Futile Recanalisation in Stroke Patients After Endovascular therapy. Implementation of the tool at https://fr-index.biomind.cn/RESCUE-FR/, where one can enter the raw information and obtain risk scores. (b) According to the risk stratification standard (low risk<35%, intermediate risk 35%–70%, high risk >70%), the classification accuracy was demonstrated in the validation cohort, RESCUE-FR, Registration study for Critical Care of Acute Ischemic Stroke: Prediction of Futile Recanalisation.
- Table 1
Important characteristics of the derivation and validation cohorts
Characteristics Derivation cohort (N=945) Validation cohort (N=177) P value Baseline characteristics Male, n (%) 605 (64.02) 122 (68.93) 0.21 Age, mean±SD, years 64.92±12.18 64.04±12.73 0.38 NIHSS, median (IQR) 15 (11–21) 14 (10–17) 0.01 Medical history, n (%) Diabetes 209 (22.12) 41 (23.16) 0.76 Hypertension 538 (56.93) 104 (58.76) 0.65 Stroke/TIA 179 (18.94) 41 (23.16) 0.19 Atrial fibrillation 79 (8.36) 26 (14.69) 0.01 Current smoke 335 (35.52) 66 (37.29) 0.65 IVT, n (%) 281 (29.74) 56 (31.64) 0.61 Occlusion location ICA 253 (26.77) 36 (20.34) 0.07 ACA 15 (1.59) 5 (2.82) 0.25 MCA 479 (50.69) 74 (41.81) 0.03 VA 84 (8.89) 7 (3.95) 0.03 BA 151 (15.98) 16 (9.04) 0.02 PCA 18 (1.90) 2 (1.13) 0.47 Time intervals, median (IQR), min Onset to door 186 (97–346) 204 (110–300) 0.01 Onset to recanalisation 480 (344–658) 472 (351–640) 0.44 Passes, n (%) 0.32 <3 593 (62.75) 118 (66.67) ≥3 352 (47.25) 59 (33.33) Post-EVT mTICI, n (%) 0.75 2b 278 (29.42) 50 (28.25) 3 667 (70.58) 127 (71.75) 24 hours follow-up, median (IQR) 24 hours GCS 11 (7–14) 11(7–15) 0.77 24 hours NIHSS 13 (7–19) 13 (6–18) 0.71 Infarct volume at 24 hours, mL 20.31 (6.82–60.38) 26.65 (8.91–97.90) 0.10 mRS 3–6 at 90 days, n (%) 510 (53.97) 107 (60.45) 0.11 ACA, anterior cerebral artery; BA, basilar artery; EVT, endovascular treatment; GCS, Glasgow Coma Scale; ICA, internal carotid artery; IVT, intravenous thrombolysis; MCA, middle cerebral artery; mRS, modified Rankin Scale; mTICI, modified Thrombolysis in Cerebral Infarction; NIHSS, National Institutes of Health Stroke Scale; PCA, posterior cerebral artery; TIA, transient ischaemic attacks; VA, vertebral artery.
- Table 2
The performance metrics of model 6 using TFSP in the derivation and validation cohorts
Data set Classifier Accuracy AUC SPE SEN Precision F1 Log Derivation RF 0.77 0.85 (0.78–0.93) 0.72 0.83 0.78 0.80 0.48 Validation RF 0.75 0.80 (0.73–0.87) 0.71 0.78 0.81 0.75 0.55 AUC, area under the curve; Log, log loss; RF, random forest; SEN, sensitivity; SPE, specificity; TFSP, two-stage feature selection pipeline.
Supplementary Materials
Supplementary data
Additional Files
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
