Functional Ultrasound Imaging Combined with Machine Learning for Whole-Brain Analysis of Drug-Induced Hemodynamic...

Area of Science:

• Neuroscience
• Pharmacology
• Medical Imaging

Background:

• Functional ultrasound imaging (fUSI) visualizes cerebral blood volume changes with high spatiotemporal resolution.
• Current fUSI studies often use predefined regions of interest, potentially missing critical off-target brain activity.
• Dizocilpine (MK-801), an NMDA receptor antagonist, is used in preclinical models to study memory and learning impairments.

Purpose of the Study:

• To compare machine learning approaches (CNN, SVM, ViT) combined with fUSI for analyzing drug pharmacodynamics.
• To develop a data-driven method for identifying and mapping drug effects in the brain.
• To investigate the effects of Dizocilpine (MK-801) on brain activity using fUSI and machine learning.

Main Methods:

• Functional ultrasound imaging (fUSI) was used to measure cerebral blood volume changes.
• Three machine learning models (CNN, SVM, ViT) were applied to fUSI data.
• Dizocilpine (MK-801) was administered to preclinical models, and brain activity was analyzed.

Main Results:

• All tested machine learning models distinguished between drug and control conditions.
• Convolutional neural networks (CNNs) demonstrated superior performance in capturing spatial features and anatomical specificity.
• Class activation mapping identified key brain regions, including the prefrontal cortex and hippocampus, affected by MK-801.

Conclusions:

• The combination of fUSI and CNN provides a powerful analytical framework for preclinical drug studies.
• This integrated approach enables data-driven identification and mapping of drug-induced brain activity.
• The findings highlight the utility of fUSI-CNN for understanding drug mechanisms while preserving anatomical and physiological context.