A multi-model framework to estimate perfusion parameters using contrast-enhanced ultrasound imaging

Area of Science:

• Medical Imaging
• Biophysics
• Quantitative Ultrasound

Background:

• Contrast-enhanced ultrasound (CEUS) enables real-time tissue perfusion imaging and quantification.
• Traditional least-squares curve fitting can be inadequate for complex, noisy perfusion data.
• Perfusion model relevance varies by organ/tissue, necessitating adaptable quantification.

Purpose of the Study:

• To develop and validate a multi-model framework for selecting optimal perfusion models and fitting methods.
• To improve the robustness and accuracy of perfusion parameter estimation in CEUS.

Main Methods:

• Utilized a system identification approach for model selection based on best fit to time-intensity curves.
• Compared the multi-model framework against single-model methods using 3D DCE-US in vivo mouse cancer models and simulation data.
• Evaluated fit quality using Spearman correlation, R², and NRMSE.

Main Results:

• The multi-model framework demonstrated superior performance over conventional single-model approaches.
• Achieved R² of 0.98 and NRMSE of 0.18, indicating robust parameter estimation.
• Traditional methods showed higher variability (R²=0.91, NRMSE=0.31) dependent on model selection.

Conclusions:

• The proposed multi-model framework provides more robust perfusion parameter estimation than single-model least-squares methods.
• The technique is minimally sensitive to noise, frame rate, and other data acquisition variables.
• This adaptable framework is valuable for modeling diverse perfusion systems across different organs and tissues.