Quantification method in [18F]fluorodeoxyglucose brain positron emission tomography using independent component...

Area of Science:

• Nuclear medicine
• Medical imaging
• Quantitative analysis

Background:

• Accurate input functions are crucial for quantitative analysis in dynamic brain positron emission tomography (DBPET).
• Traditional methods rely on blood sampling, which can be invasive and introduce errors.
• Independent Component Analysis (ICA) offers a potential non-invasive alternative for deriving input functions.

Purpose of the Study:

• To develop and validate an accurate image-derived input function (IDIF) extraction method using ICA for DBPET.
• To compare the performance of the ICA-based method against conventional blood sampling techniques.
• To assess the accuracy of quantitative analysis, specifically the metabolic rate of [F]-2-fluoro-2-deoxy-D-glucose (MRFDG), using different input function estimation methods.

Main Methods:

• A modified linear model with haematocrit correction was applied to enhance ICA-based input function estimation.
• The methodology was tested using both simulated dynamic phantoms and clinical 2-hour DBPET scans.
• Image-derived input functions from ICA (Iica) and blood sampling (Itp) were compared, along with a reference input function from late blood samples (Iest).

Main Results:

• Simulated studies showed lower mean percentage errors for MRFDG with ICA (4.2%) compared to the reference method (8.2%).
• Clinical studies revealed that the ICA-based input function (Iica) yielded a mean percentage error of 7.7% ± 3.3% for MRFDG, outperforming the reference method (12.6% ± 7.5%).
• The ICA-derived input functions demonstrated high correlation with gold-standard blood sampling results.

Conclusions:

• A novel technique for estimating image-derived input functions using ICA without the need for blood sampling has been successfully developed.
• The proposed ICA method provides accurate input functions for DBPET, comparable to traditional blood sampling.
• This non-invasive approach offers improved accuracy and reduced error in quantitative PET analyses, surpassing previously reported methods.