Parkinson's Disease STN-DBS Patient Data for Predicting Postoperative FOGQ Worsening

A multimodal dataset from 134 Parkinson's disease patients undergoing bilateral subthalamic nucleus deep brain stimulation (STN-DBS). The data integrates clinical scale assessments, structural neuroimaging features, stereotactic electrode localization data, and radiomics features from preoperative MRI. It was used to develop and validate a LightGBM machine learning model achieving an AUC of 0.917 for predicting postoperative worsening of freezing of gait questionnaire scores.

Use Cases

• Predict postoperative gait deterioration in Parkinson's patients based on integrated clinical and imaging features.
• Evaluate the predictive importance of different data modalities (clinical, structural, radiomic) using SHAP value analysis.
• Benchmark machine learning algorithms for clinical outcome prediction using a nested cross-validation framework.
• Support preoperative patient selection and treatment planning for STN-DBS surgery.

Strengths

• Integrates four distinct data modalities: clinical scales, structural neuroimaging, electrode localization, and radiomics.
• Model development used a rigorous 5-fold nested cross-validation framework.
• The resulting LightGBM model achieved a high AUC of 0.917 for the prediction task.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The study is retrospective; external validation in prospective multicenter cohorts is noted as a necessary next step.

Provenance

Source

figshare

Collection Method

Retrospective analysis of patient data.

Freshness

Last updated 2026-05-05 04:10:28; freshness should be verified.