Table 1_GLOBE: an explainable machine learning platform for preoperative prediction of thr

427 patients with glioma were included in a retrospective cohort study to develop machine learning models for predicting postoperative venous thromboembolism (VTE) and postoperative neurological deterioration (PND). The models, including a Top-10 random forest for VTE and a Top-10 logistic regression for PND, were validated on a held-out test set and deployed in an online platform by Feiling Xiang. The dataset was last updated on June 1, 2026.

Use Cases

• Predicting postoperative venous thromboembolism risk based on patient age, D-dimer, fibrin degradation products, and hemoglobin levels.
• Predicting postoperative neurological deterioration risk based on patient age, glioma recurrence status, Caprini score, neutrophil percentage, and hypertension history.
• Benchmarking machine learning models against conventional clinical scoring systems like the Caprini score.
• Developing interpretable clinical tools using SHAP analysis for feature importance.
• Integrating predictive models into a web-based platform for real-time preoperative risk assessment.

Strengths

• Model performance for VTE prediction achieved an AUC of 0.815 in the test set, with a clinically significant sensitivity analysis showing an AUC of 0.923.
• The dataset underpins a deployed, accessible online prediction platform (GLOBE) for real-time use.
• Models were developed using a cohort of 427 patients with specific, clinically relevant outcome rates (34% VTE, 35% PND).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small at 28.2 KB, indicating limited scope.

Provenance

Source

Feiling Xiang

Collection Method

Retrospective cohort study of 427 patients with glioma.

Freshness

Last updated 2026-06-01 05:25:30; freshness should be verified.