Cardiothoracic & vascular anesthesia
Functional MRI-based machine learning strategy for prediction of postoperative delirium in cardiac surgery patients: A secondary analysis of a prospective observational study
Authors' conclusion
This study found that the random forest model using preoperative functional MRI data and clinical variables was accurate in identifying patients at high risk of developing delirium after cardiac surgery.
Study context
- Population
- Patients undergoing preoperative functional MRI scan and cardiac valve replacement at one tertiary hospital in China
- Participants
- 103 patients
- Comparison
- Random forest, logistic regression and linear support vector machine models trained on preoperative resting-state fMRI functional connectivity (FC) measures (± clinical variables) to predict postoperative delirium
Methodological detailssingle-centre, machine learning prediction model
- Single-centre
- Machine learning prediction model
Primary result
As reported in the abstract
Postoperative delirium prediction performance
| Group | AUC | Sensitivity | Specificity | Accuracy | Precision | F1 score |
|---|---|---|---|---|---|---|
| Random forest, degree of FC | 0.924 | — not stated in the abstract | 0.894 | 0.864 | 0.887 | 0.859 |
| Random forest, degree of FC + clinical variables | 0.928 | 0.909 | — not stated in the abstract | 0.879 | — not stated in the abstract | 0.882 |
AUC
Random forest, degree of FC0.924
Random forest, degree of FC + clinical variables0.928
Sensitivity
Random forest, degree of FC— not stated in the abstract
Random forest, degree of FC + clinical variables0.909
Specificity
Random forest, degree of FC0.894
Random forest, degree of FC + clinical variables— not stated in the abstract
Accuracy
Random forest, degree of FC0.864
Random forest, degree of FC + clinical variables0.879
Precision
Random forest, degree of FC0.887
Random forest, degree of FC + clinical variables— not stated in the abstract
F1 score
Random forest, degree of FC0.859
Random forest, degree of FC + clinical variables0.882
— not stated in the abstract
Further findings
Incidence of postoperative delirium
- Participants meeting postoperative delirium criteria
- 29 (28.2 %)
Most discriminative nodes
- Located in the default, cingulo-opercular and frontoparietal networks
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Functional MRI-based machine learning strategy for prediction of postoperative delirium in cardiac surgery patients: A secondary analysis of a prospective observational study. J Clin Anesth. Published 13 February 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →