A multimodal image feature extraction method for x-ray grating phase contrast computed tomography based on monogenic...

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Processing

Background:

• Computed tomography (CT) using X-ray absorption imaging is crucial in clinical medicine.
• X-ray phase contrast imaging (XPCI) is emerging, offering separation of attenuation, refraction, and scattering signals.
• Existing CT image recognition methods primarily focus on absorption contrast (AC-CT) images, neglecting other XPCI-CT modalities.

Purpose of the Study:

• To introduce a novel method for enhancing feature recognition in XPCI-CT images.
• To address the limitations of existing algorithms in handling various XPCI-CT contrast images.
• To improve the quality and interpretability of XPCI-CT data.

Main Methods:

• Development of the variable kernel multi-scale adaptive monogenic signal phase consistency (VK-MA PC) model.
• Construction of monogenic signals using filters tailored to the characteristics of different contrast images (absorption, differential phase, dark field).
• Application of multi-scale analysis and optional pre-decomposition for enhanced image feature extraction.

Main Results:

• The VK-MA PC model demonstrates improved image feature extraction capabilities.
• Experimental validation on 4D extended cardiac-torso (XCAT) human body simulation data.
• Successful application to laboratory fish XPCI-CT data, showing potential applicability.

Conclusions:

• The VK-MA PC model offers a promising approach for feature recognition in XPCI-CT.
• The method effectively handles diverse contrast images generated by XPCI-CT.
• This advancement has significant potential for the field of XPCI-CT analysis.