Development and Validation of a Machine Learning Model for Dementia Staging in a Heterogeneous Cognitive Impairment...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• Precise dementia staging is crucial for emerging disease-modifying therapies.
• Objective staging methods are needed to complement current clinical assessments.

Purpose of the Study:

• To develop and validate a machine learning model for objective dementia severity staging.
• Integrate multimodal data including demographics, neuropsychological scores, and PET imaging.
• Enhance patient management and candidate screening for novel therapies.

Main Methods:

• Recruited 149 patients (100 with Alzheimer's disease).
• Utilized demographic, neuropsychological, and multimodal PET data.
• Trained seven machine learning algorithms, with XGBoost showing superior performance.
• Assessed model interpretability using SHAP analysis.

Main Results:

• The XGBoost model achieved an AUC of 0.888 in the validation cohort.
• Key predictors included education, MMSE, and composite SUVR scores (amyloid and FDG).
• SHAP analysis highlighted MMSE and FDG SUVR as significant contributors.

Conclusions:

• An interpretable multimodal machine learning model for dementia staging was developed and internally validated.
• The model shows potential for optimizing patient management and therapy candidate screening.
• Further external validation is required for clinical application.