Stepwise local stitching ultrasound image algorithms based on adaptive iterative threshold Harris corner features.

Area of Science:

• Medical Imaging
• Computer Vision
• Ultrasound Technology

Background:

• Accurate delineation of organs at risk (OARs) is crucial for effective radiation therapy planning in pelvic cancers.
• Current imaging techniques may have limitations in providing comprehensive views for precise OAR segmentation.
• Ultrasound (US) imaging offers real-time visualization but requires advanced algorithms for stitching and feature extraction.

Purpose of the Study:

• To propose and evaluate a novel Harris corner detection algorithm, Adaptive Iterative Threshold-Harris (AIT-Harris), for improved wide-field ultrasound image generation.
• To compare the performance of AIT-Harris with traditional Harris and Morave algorithms in feature extraction for image stitching.
• To assess the accuracy of OAR contouring using stitched wide-field US images compared to Cone-beam Computed Tomography (CBCT).

Main Methods:

• A new AIT-Harris algorithm was developed, integrating iterated threshold segmentation with adaptive iterative thresholding for corner detection.
• A stepwise local stitching algorithm was employed to construct wide-field US images using extracted corner features.
• Corner matching rates were compared across AIT-Harris, Harris, and Morave algorithms using paired sample t-tests.
• The accuracy of OAR delineation on stitched US images was evaluated against CBCT, calculating Dice and Jaccard similarity coefficients.

Main Results:

• The AIT-Harris algorithm demonstrated statistically significant improvements in corner matching rates compared to the Morave and Harris algorithms (P < .05).
• Stitched wide-field US images enabled accurate delineation of OARs, with an average Dice similarity coefficient of 0.924 and Jaccard coefficient of 0.894 for the bladder.
• The proposed method enhanced the accuracy of corner detection, leading to improved wide-field US image quality for OAR delineation.

Conclusions:

• The AIT-Harris algorithm provides a more accurate method for corner detection in ultrasound image processing.
• Stitched wide-field ultrasound images generated using AIT-Harris can effectively modify the delineation range of OARs in the pelvic cavity.
• This advancement holds potential for improving radiotherapy planning and precision in gynecologic and prostate cancer treatments.