Hemodynamic segmentation of brain perfusion images with delay and dispersion effects using an...

Area of Science:

• Medical Imaging
• Neuroscience
• Biomedical Engineering

Background:

• Dynamic susceptibility contrast (DSC) MRI is crucial for diagnosing cerebrovascular diseases.
• Segmentation of perfusion compartments aids clinical diagnosis and treatment.
• Cerebrovascular diseases can alter hemodynamic signals, challenging current segmentation accuracy.

Purpose of the Study:

• To improve the accuracy of automatic perfusion compartment identification in brain MRI.
• To assess segmentation technique performance under conditions of delayed and dispersed perfusion.
• To enhance the evaluation of cerebrovascular disease severity.

Main Methods:

• Improved expectation-maximization algorithm using hierarchical clustering results as initial parameters.
• Developed a mixture of multivariate Gaussians model for segmentation.
• Utilized Monte Carlo simulations to evaluate performance under varying signal profile conditions (delay, dispersion, noise).

Main Results:

• The proposed method successfully differentiated normal, delayed, and dispersed hemodynamics in patient data.
• Accurate identification of local arterial input function for impaired tissues was achieved.
• Minimized errors in cerebral blood flow estimation.
• Successfully identified tissues at risk of infarct and those with or without collateral blood supply.

Conclusions:

• The enhanced segmentation method improves the accuracy of perfusion analysis in cerebrovascular diseases.
• This technique aids in precise disease severity evaluation and risk stratification.
• Enables better identification of impaired tissue and collateral circulation, crucial for treatment planning.