A Signal-Domain Object Segmentation Method for Ultrasound and Photoacoustic Computed Tomography.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Imaging

Background:

• Image segmentation is crucial for enhancing diagnostic capabilities in ultrasound computed tomography (USCT) and photoacoustic computed tomography (PACT).
• Traditional image domain segmentation methods struggle with low contrast, noise, and artifacts common in reconstructed images.
• Existing methods often require image reconstruction prior to segmentation, adding complexity and potential for error.

Purpose of the Study:

• To introduce a novel, automatic signal domain object segmentation method for USCT and PACT.
• To overcome the limitations of image domain segmentation by performing segmentation before image reconstruction.
• To develop a computationally efficient, accurate, and robust segmentation technique.

Main Methods:

• Established a relationship between the time-of-flight (TOF) of received waves and the object's boundary, modeled by ellipse equations.
• Utilized common tangents of neighboring ellipses to identify tangent points approximating the object boundary with high fidelity.
• Developed a signal domain approach that bypasses the need for initial image reconstruction.

Main Results:

• The proposed method achieved automatic, robust, and computationally efficient segmentation.
• Experimental results on human fingers and mice cross-sections demonstrated segmentation accuracy equivalent or superior to active contour methods.
• The signal domain approach successfully segmented objects without requiring prior image reconstruction.

Conclusions:

• The new signal domain segmentation method significantly reduces complexity in USCT and PACT.
• The technique shows great potential for eliminating user dependency while maintaining high segmentation accuracy.
• This method can be seamlessly integrated into other USCT and PACT processing algorithms, including image reconstruction.