Robust estimation of the apparent diffusion coefficient (ADC) in heterogeneous solid tumors.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Biomedical Engineering
• Medical Physics

Background:

• The least-squares algorithm is commonly used for estimating apparent diffusion coefficient (ADC) from MRI magnitude data.
• This method is often assumed to have a negligible bias, but its impact on tumor ADC estimation is not fully understood.

Purpose of the Study:

• To evaluate the in vivo effect of least-squares bias on tumor ADC estimates.
• To compare the accuracy of least-squares versus a robust maximum likelihood approach for ADC estimation in tumors.
• To investigate the influence of ADC, signal-to-noise ratio (SNR), and tissue heterogeneity on estimation bias.

Main Methods:

• In vivo evaluation of tumor ADC estimation using least-squares and maximum likelihood approaches.
• Monte Carlo simulations to analyze the bias magnitude under varying ADC and SNR conditions.
• Assessment of the ability to resolve necrotic from viable tumor regions with each method.

Main Results:

• The least-squares algorithm consistently and significantly underestimates tumor ADC values by an average of 23.4 +/- 12% compared to the maximum likelihood approach.
• Bias magnitude increases with higher ADC values and lower signal-to-noise ratios (SNR).
• In vivo, least-squares resulted in reduced ability to differentiate necrotic from viable tumor tissue.

Conclusions:

• The least-squares algorithm introduces a significant bias in tumor ADC estimation from magnitude MR data.
• A robust maximum likelihood approach is recommended over least-squares for analyzing diffusion MRI data, especially in heterogeneous tumors.
• Accurate ADC estimation is critical for evaluating tumor characteristics and treatment response, particularly with advanced diffusion models and modest SNR data.