Use of Artificial Intelligence in Imaging Dementia

Area of Science:

• Neurology and Medical Imaging
• Artificial Intelligence in Healthcare
• Neurodegenerative Disease Diagnostics

Background:

• Alzheimer's disease is the leading cause of dementia in older adults, with diagnostic delays and misdiagnosis posing significant challenges.
• Neuroimaging is crucial for diagnosing neurodegenerative diseases but is susceptible to human error due to data complexity.
• There is a critical need for improved diagnostic accuracy and efficiency in dementia patient care.

Purpose of the Study:

• To evaluate the potential of artificial intelligence (AI) algorithms, specifically graph convolutional networks and convolutional neural networks, in automating neuroimaging interpretation for dementia diagnosis.
• To assess the impact of AI assistance on radiologist performance in detecting specific imaging abnormalities, such as amyloid-related imaging abnormalities (ARIA-E and ARIA-H).
• To explore the feasibility of translating AI methods into clinical practice for improved diagnosis and prognosis of neurodegenerative diseases.

Main Methods:

• Utilized graph convolutional network (GCN) frameworks for multimodal sparse interpretability in detecting Alzheimer's disease and mild cognitive impairment.
• Developed and validated a convolutional neural network (CNN) model using FDG-PET scans to predict clinical diagnoses including Alzheimer's disease, dementia with Lewy bodies, and mild cognitive impairment.
• Compared radiologist detection performance for ARIA-E and ARIA-H with and without AI assistance.

Main Results:

• AI-assisted interpretation significantly improved radiologist sensitivity in detecting ARIA-E (87% vs. 71%) and ARIA-H (79% vs. 69%) compared to unassisted interpretation.
• The developed CNN model demonstrated predictive accuracy for final clinical diagnoses based on FDG-PET imaging.
• GCN frameworks showed promise in supporting the detection of Alzheimer's disease and its prodromal stage, mild cognitive impairment.

Conclusions:

• AI-driven automation of neuroimaging interpretation holds significant potential to reduce diagnostic errors and bias in dementia assessment.
• AI tools can enhance clinical decision-making and improve diagnostic accuracy, particularly in identifying subtle imaging biomarkers.
• Despite implementation challenges, AI in neuroimaging offers a transformative approach to dementia diagnosis and patient management, promising improved health outcomes.