Application of a virtual imaging framework for investigating a deep learning-based reconstruction method for 3D...

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Imaging

Background:

• Quantitative photoacoustic computed tomography (qPACT) shows promise for physiological parameter estimation, like blood oxygen saturation.
• Robust qPACT reconstruction faces challenges from computational load, modeling complexity, and experimental variability.
• Learning-based methods offer potential solutions but require rigorous validation.

Purpose of the Study:

• To develop and utilize a realistic virtual imaging (VI) testbed for validating 3D learning-based qPACT reconstruction methods.
• To assess the performance of a representative learning-based qPACT method for breast imaging applications.
• To evaluate the method's robustness against subject variability and physical factors like noise and acoustic aberrations.

Main Methods:

• Development of a realistic virtual imaging testbed using stochastically generated numerical phantoms reflecting anatomical and physiological variations.
• Application of the VI testbed to evaluate a 3D learning-based qPACT reconstruction algorithm for breast imaging.
• Systematic assessment of the reconstruction method's performance under varying conditions, including measurement noise and acoustic aberrations.

Main Results:

• The realistic VI testbed enabled a comprehensive evaluation of a 3D learning-based qPACT reconstruction method.
• Performance insights were gained regarding the method's strengths and limitations across different subject variations and physical factors.
• The study highlights the importance of realistic VI environments for validating advanced imaging reconstruction techniques.

Conclusions:

• A realistic VI testbed is crucial for the robust validation of learning-based qPACT reconstruction methods.
• The employed learning-based method shows potential for breast imaging qPACT but requires careful consideration of its limitations.
• This work establishes a framework for future validation studies of qPACT reconstruction algorithms.