Voxel-accurate MRI-microscopy Correlation Enables AI-powered Prediction of Brain Disease States

Area of Science:

• Neuroimaging
• Radiology
• Computational Biology

Background:

• Magnetic resonance imaging (MRI) is crucial for brain visualization but lacks cellular-level understanding of signal origins and temporal changes.
• Current MRI techniques struggle to precisely link imaging signals to underlying biological structures and processes.
• Understanding the cellular basis of MRI signals is vital for accurate diagnosis and treatment monitoring.

Purpose of the Study:

• To develop a platform for precise, voxel-level mapping of MRI signals to their biological ground truth.
• To enable longitudinal, in vivo correlation of MRI findings with cellular and microenvironmental details.
• To leverage this integrated data for training deep learning models to enhance MRI resolution and interpretation.

Main Methods:

• Introduction of BRIDGE (Brain Radiological Imaging with Deep-learning based Ground-Truth Exploration), an integrated platform.
• Multi-step, iterative co-registration pipeline linking in vivo MRI with in vivo two-photon and ex vivo super-resolution microscopy.
• Development of convolutional neural networks for enhancing MRI effective resolution using registered data.

Main Results:

• Identification of a microenvironmental vessel biomarker for early metastatic colonization in breast cancer brain metastasis models.
• Correlation of T2*-weighted hypointense lesions with reduced blood flow and erythrostasis in perimetastatic capillaries.
• Demonstration of direct correlations between T2-weighted signal changes and tumor growth dynamics in glioma models.

Conclusions:

• BRIDGE establishes a microscopic ground truth for MRI signatures, advancing radiological interpretation.
• The platform enables deep learning-based predictive histology and provides cellular insights into tumor microenvironments.
• This approach offers direct clinical imaging implications for understanding and diagnosing brain diseases.