Development of a Postoperative Delirium Risk Prediction Model for Elderly Patients by Integrating MRI Radiomics With...

Area of Science:

• Geriatric Medicine
• Radiology
• Neuroscience

Background:

• Postoperative delirium (POD) is a common complication in elderly surgical patients.
• Early identification and prevention of POD are crucial for improving patient outcomes.
• Current risk assessment methods may not fully capture the complex factors contributing to POD.

Purpose of the Study:

• To evaluate the efficacy of magnetic resonance imaging (MRI) radiomics in assessing POD risk in elderly patients.
• To develop an integrated predictive model for POD by combining radiomic and clinical data.
• To provide a clinical decision support tool for identifying high-risk patients.

Main Methods:

• Prospective cohort study of 300 elderly patients (≥65 years) undergoing elective surgery.
• MRI features (hippocampal volume, intracranial volume, white matter hyperintensity) and clinical data were collected.
• A predictive model was constructed and validated using statistical analysis (R 4.2.2).

Main Results:

• Several factors, including age, surgical duration, and MRI-derived volumes, correlated with POD.
• Multivariate analysis identified age, surgical duration, hippocampal volume, and white matter hyperintensity as independent risk factors.
• The integrated prediction model demonstrated high predictive accuracy (ROC AUCs of 0.973 and 0.966).

Conclusions:

• The developed nomogram prediction model effectively predicts POD risk in elderly patients.
• This model can serve as a valuable clinical decision support tool.
• Implementing this tool may help reduce POD incidence through targeted preventive measures.