Positional Prompts-Enhanced Brain-Heart-Gut Interactions for Mild Cognitive Impairment Diagnosis.

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Mild cognitive impairment (MCI) is an early stage of dementia linked to complex brain and peripheral organ interactions.
• Heart dysfunction and gut microbiota imbalances are increasingly recognized as contributors to MCI.
• Current diagnostic approaches for MCI do not fully leverage cross-organ interactions.

Purpose of the Study:

• To propose a novel diagnostic framework for MCI that integrates information from the brain, heart, and gut.
• To develop a method for MCI diagnosis using whole-body PET images, applicable even when only brain MRI or PET data is available.
• To enhance MCI diagnosis by modeling inter-organ crosstalk.

Main Methods:

• Collected a large, multi-modal dataset including whole-body PET, brain MRI, and brain PET images from multiple centers.
• Developed organ-specific image encoders and utilized hierarchical Transformers to model brain-heart, brain-gut, and brain-heart-gut interactions.
• Implemented a multi-level knowledge distillation scheme to transfer a comprehensive brain-heart-gut model to a brain-only diagnostic model.

Main Results:

• The proposed framework effectively integrates brain, heart, and gut features for improved MCI diagnosis.
• Demonstrated superior performance compared to state-of-the-art methods on multi-center data.
• Successfully adapted the model for MCI diagnosis using only brain imaging data.

Conclusions:

• The integration of heart and gut interactions significantly enhances MCI diagnosis.
• The novel framework offers a promising approach for early detection and diagnosis of MCI.
• This work highlights the potential of leveraging whole-body imaging and AI for neurodegenerative disease diagnostics.