CPSN: Caputo Principal-Curve-Guided Segmentation Network on Ultrasound Kidney Databases.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Nephrology

Background:

• Accurate kidney contour segmentation is crucial for surgical planning and anatomical modeling.
• Manual segmentation is time-consuming and suffers from observer variability.
• Ultrasound kidney boundary delineation is challenging due to indistinct boundaries.

Purpose of the Study:

• To develop a novel, precise method for kidney boundary delineation in ultrasound images.
• To overcome the limitations of manual segmentation and indistinct boundaries in ultrasound data.
• To enhance the accuracy of kidney segmentation for clinical applications.

Main Methods:

• Developed a Caputo principal-curve-guided segmentation network (CPSN).
• Integrated principal-curve (PC)-based vertex decision into a Caputo multiple-layer learning network.
• Employed an initial deep network for rough contour extraction and a Caputo training network for error reduction and alignment.

Main Results:

• The CPSN achieved superior performance compared to state-of-the-art techniques on multi-institute ultrasound kidney databases.
• Achieved high accuracy with Dice index (DI) of 94.6±3.2%, Jaccard index (JI) of 93.4±3.7%, and accuracy (ACC) of 94.1±3.47%.

Conclusions:

• The proposed CPSN method offers a robust and effective solution for ultrasound kidney segmentation.
• This automated approach can significantly improve precision in pre-operative planning and image-guided procedures.
• The method demonstrates high accuracy and robustness across diverse ultrasound datasets.