Simulation of realistic abnormal SPECT brain perfusion images: application in semi-quantitative analysis

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Functional image analysis methods require thorough validation, which is often lacking.
• Current simulation techniques for medical imaging suffer from long run times and unrealistic outputs.
• Developing robust simulation methods is crucial for validating brain imaging analysis tools.

Purpose of the Study:

• To introduce an efficient method for simulating realistic brain Single-Photon Emission Computed Tomography (SPECT) images with known abnormalities.
• To generate a dataset for the UK audit of SPECT image analysis methods.
• To enable the cross-validation and optimization of semi-quantitative analysis techniques in brain SPECT.

Main Methods:

• Simulating abnormalities in normal brain SPECT images using a measured point spread function (PSF) and a stereotactic atlas.
• Defining abnormalities in stereotactic space, transforming them to subject space, and applying PSF convolution.
• Analyzing simulated datasets using SPM99 and the MarsBaR VOI toolbox, comparing results against ground truth.

Main Results:

• Using a smoothing kernel equal to system resolution improved analysis compared to larger kernels.
• A significant correlation was observed between simulated abnormality volume and SPM99 detected size.
• Region median provided better sensitivity in VOI analysis than region mean.

Conclusions:

• The proposed method offers an efficient approach for simulating brain SPECT abnormalities.
• The generated dataset facilitates the comparison and cross-validation of semi-quantitative analysis methods.
• This methodology aids in optimizing analysis parameters for improved accuracy in brain SPECT imaging.