Lesion segmentation for MR spectroscopic imaging using the convolution difference method.

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Medical Physics

Background:

• Volumetric Magnetic Resonance Spectroscopy (MRSI) provides metabolic information within brain lesions.
• Accurate segmentation of lesions from MRSI metabolite ratio maps is crucial for diagnosis and treatment monitoring.
• Existing segmentation methods struggle with spatial response function variations and signal heterogeneity.

Purpose of the Study:

• To develop and evaluate a novel, robust method for delineating lesion boundaries in volumetric MRSI metabolite ratio maps.
• To account for spatial response function variations and spectral quality in lesion segmentation.
• To compare the performance of the new method against existing thresholding techniques.

Main Methods:

• A novel 'convolution difference' method was developed for lesion segmentation.
• Procedures were established for processing metabolite ratio maps and excluding low-quality spectral regions.
• The convolution difference method was evaluated using computer simulations and in vivo MRSI studies of gliomas, comparing it to iterative and fixed amplitude thresholding techniques.

Main Results:

• Computer simulations demonstrated superior performance of the convolution difference method for segmenting ratio maps.
• In vivo studies showed variations in tumor volume measurements between the convolution difference and iterative thresholding methods.
• Both novel methods exhibited improved accuracy compared to using a fixed amplitude threshold in visual analysis.

Conclusions:

• Fixed thresholding methods can lead to significant errors in lesion volume definition, especially with broad spatial response functions.
• The novel convolution difference method offers a robust approach for segmenting volumetric MRSI metabolite data.
• This technique enhances the accuracy of lesion boundary delineation in neuroimaging studies.