Ultrasound Prostate Segmentation Using Adaptive Selection Principal Curve and Smooth Mathematical Model

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Accurate prostate segmentation in ultrasound images is vital for diagnosing prostate cancer and guiding surgery.
• Ultrasound images present segmentation challenges due to low signal-to-noise ratio, poor contrast, and indistinct prostate boundaries.

Purpose of the Study:

• To develop a novel hybrid method for accurate prostate segmentation in ultrasound images.
• To generate precise prostate contours from diverse ultrasound datasets.

Main Methods:

• A hybrid approach combining a principal curve-based method with a fractional-order neural network.
• The method utilizes projection indices from principal curves as input for the neural network.
• A smooth mathematical map is generated by the neural network to represent the prostate boundary.

Main Results:

• The proposed hybrid method demonstrated superior performance compared to state-of-the-art segmentation techniques.
• The method achieved accurate prostate segmentation across multiple datasets from various institutions.
• Experimental evaluations confirmed the robustness of the method across different evaluation metrics.

Conclusions:

• The novel hybrid method offers a robust and accurate solution for prostate segmentation in ultrasound imaging.
• This technique has significant potential for improving prostate cancer diagnosis and image-guided surgical procedures.