Machine Learning Model to Predict Diagnosis of Mild Cognitive Impairment by Using Radiomic and Amyloid Brain PET

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Biomarkers

Background:

• Mild cognitive impairment (MCI) is a precursor to Alzheimer's disease.
• Early and accurate diagnosis of MCI is crucial for timely intervention.
• Current diagnostic methods have limitations in sensitivity and specificity.

Purpose of the Study:

• To develop a deep learning model for predicting amnestic MCI (aMCI) diagnosis.
• To utilize radiomic features and amyloid positron emission tomography (PET) data.
• To compare the model's performance against standard SUV measurements.

Main Methods:

• A cohort of 328 subjects (normal controls and aMCI) from ADNI and a clinical trial were analyzed.
• Radiomic features and amyloid loads were extracted from brain PET scans.
• A feed-forward neural network was trained and validated using selected features.

Main Results:

• The deep learning model achieved an Area Under the Curve (AUC) of 90% for aMCI prediction.
• The model demonstrated 80% accuracy and 78% F1-score on the test set.
• This outperformed SUV performance, which had an AUC of 71%.

Conclusions:

• The developed machine learning model accurately identifies MCI subjects with 84% specificity and 81% sensitivity.
• Deep learning algorithms integrating radiomic data and amyloid PET load enhance MCI diagnosis prediction.
• This approach offers a significant improvement over using SUV alone for MCI diagnosis.