Quantitative body DW-MRI biomarkers uncertainty estimation using unscented wild-bootstrap.

Area of Science:

• Medical Imaging
• Biophysics
• Quantitative MRI

Background:

• Accurate estimation of diffusion parameters from diffusion-weighted magnetic resonance imaging (DW-MRI) is crucial for assessing pathology in clinical applications.
• Traditional uncertainty estimation methods often require repeated image acquisitions, which is impractical for routine clinical use, especially in body imaging.
• Existing model-based bootstrap techniques are unsuitable for non-linear body diffusion models, limiting their application.

Purpose of the Study:

• To develop and validate a novel, non-invasive method for estimating the uncertainty of diffusion parameters in quantitative body DW-MRI.
• To address the limitations of traditional methods by accommodating the non-linear nature of body diffusion models.

Main Methods:

• The proposed method utilizes the Unscented transform to calculate residuals rescaling parameters from the non-linear body diffusion model.
• The wild-bootstrap method is then applied to infer the uncertainty of the body diffusion parameters.
• The technique was validated using both phantom experiments and human subject data.

Main Results:

• The novel method successfully identified regions with higher uncertainty in body DW-MRI model parameters.
• Validation experiments demonstrated the method's accuracy, with a relative error of -36% in the estimated uncertainty values.
• This approach provides a more reliable way to quantify uncertainty without the need for repeated acquisitions.

Conclusions:

• The developed Unscented transform and wild-bootstrap method offers an effective solution for uncertainty estimation in quantitative body DW-MRI.
• This technique enhances the clinical utility of DW-MRI by providing reliable parameter uncertainty assessment.
• The findings pave the way for more robust quantitative analysis of diffusion parameters in body imaging applications.