Deep learning-based binary classification of beta-amyloid plaques using 18 F florapronol PET.

Area of Science:

• Neuroimaging
• Artificial Intelligence in Medicine
• Alzheimer's Disease Diagnostics

Background:

• Alzheimer's disease diagnosis relies on identifying amyloid plaque deposition.
• Positron Emission Tomography (PET) imaging is crucial for visualizing amyloid plaques.
• Accurate classification of PET images is essential for timely diagnosis and treatment.

Purpose of the Study:

• To investigate the efficacy of a deep learning model, specifically a convolutional neural network (CNN), for classifying amyloid plaque deposition in brain PET images.
• To assess the CNN model's ability to differentiate between amyloid-positive and amyloid-negative cases in patients suspected of Alzheimer's disease.

Main Methods:

• Retrospective analysis of brain amyloid 18F-florapronol PET/CT images from 175 patients (2019-2022).
• Visual assessment by nuclear medicine specialists classified images as positive or negative.
• A CNN model was trained and evaluated using stratified 5-fold cross-validation and a held-out test set, with data augmentation via image rotation.

Main Results:

• The CNN model achieved an average accuracy of 0.917 ± 0.027 across cross-validation folds.
• In the testing set, the model demonstrated an accuracy of 0.914 and an Area Under the Curve (AUC) of 0.958.
• These results indicate the model's high reliability in distinguishing between amyloid-positive and negative cases.

Conclusions:

• The developed CNN model shows significant potential for accurately classifying amyloid positivity in brain PET images.
• This deep learning approach can serve as a valuable supplementary tool to improve the accuracy of clinical diagnoses for Alzheimer's disease.
• Further validation could integrate this AI tool into routine diagnostic workflows.