Clinical Performance Evaluation of an Artificial Intelligence-Powered Amyloid Brain PET Quantification Method.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Amyloid PET imaging is crucial for diagnosing Alzheimer's disease.
• Accurate quantification of amyloid uptake is essential for reliable diagnosis.
• Current software tools require rigorous clinical performance evaluation.

Purpose of the Study:

• To assess the clinical performance of BTXBrain-Amyloid, an AI software for amyloid PET image quantification.
• To compare BTXBrain-Amyloid with Statistical Parametric Mapping (SPM) in terms of accuracy and spatial normalization.
• To evaluate the software's potential as a clinical tool for amyloid PET image analysis.

Main Methods:

• 150 amyloid brain PET images were analyzed using BTXBrain-Amyloid and SPM.
• Expert visual assessment categorized images as negative or positive for amyloid uptake.
• Receiver operating characteristic (ROC) and precision-recall (PR) analyses were performed.
• Spatial normalization (SN) performance was evaluated using mutual information (MI).

Main Results:

• BTXBrain-Amyloid showed superior performance over SPM, with higher Area Under the Curve (AUC) in ROC (0.979 vs. 0.959) and PR (0.983 vs. 0.949) analyses.
• BTXBrain-Amyloid achieved higher sensitivity (0.983 vs. 0.917) and comparable specificity (0.921 vs. 0.921).
• Improved spatial normalization (MI: 0.848 vs. 0.823) and enhanced detection of deep brain differences were observed with BTXBrain-Amyloid.

Conclusions:

• BTXBrain-Amyloid demonstrates superior clinical performance compared to SPM for amyloid PET image quantification.
• The AI software offers improved accuracy, spatial normalization, and detection capabilities.
• BTXBrain-Amyloid shows significant potential as a valuable tool for clinical amyloid PET image evaluation.