Evaluating Estimation Techniques in Medical Imaging Without a Gold Standard: Experimental Validation.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Statistical Modeling

Background:

• Imaging techniques are crucial for estimating parameters like cardiac ejection fraction (EF).
• Evaluating these estimation methods is challenging in clinical practice due to the absence of a true gold standard.
• Current evaluation relies on pseudo-gold standards, comparing against accepted methods.

Purpose of the Study:

• To develop and validate a maximum-likelihood approach for comparing estimation methods directly against a gold standard.
• To overcome the limitations of pseudo-gold standard evaluations in medical imaging.
• To assess the accuracy and precision of the proposed method in real-world imaging scenarios.

Main Methods:

• Developed a maximum-likelihood statistical framework for method comparison.
• Conducted simulation studies to assess parameter estimation accuracy without a gold standard.
• Designed a physical phantom experiment using SPECT and CT imaging for volume estimation validation.

Main Results:

• Simulation studies demonstrated precise and accurate parameter estimation.
• The maximum-likelihood method successfully compared estimation techniques without a gold standard.
• Experimental validation using SPECT and CT provided further evidence of the method's efficacy.

Conclusions:

• The proposed maximum-likelihood approach offers a robust alternative for evaluating imaging estimation methods.
• This method eliminates the need for a gold standard, simplifying and improving evaluation accuracy.
• The technique holds significant potential for advancing quantitative imaging analysis in clinical and research settings.