Related Experiment Video
Updated: Dec 25, 2025
08:40
Quantitation of Protein Expression and Co-localization Using Multiplexed Immuno-histochemical Staining and Multispectral Imaging
Published on: April 8, 2016
13.3K
Multi-atlas-based auto-segmentation for prostatic urethra using novel prediction of deformable image registration
Hisamichi Takagi1, Noriyuki Kadoya2, Tomohiro Kajikawa2
1Course of Radiological Technology, Health Sciences, Tohoku University Graduate School of Medicine, Sendai, Miyagi, 980-8575, Japan.
Medical Physics
|March 24, 2020
Summary
A new machine learning method accurately identifies the prostatic urethra for prostate cancer radiation therapy. This approach improves accuracy and could eliminate the need for urinary catheters during intensity-modulated radiation therapy (IMRT) planning.
Area of Science:
- Medical Imaging
- Radiation Oncology
- Machine Learning
Background:
- Accurate identification of the prostatic urethra is crucial for precise radiation dosing and toxicity evaluation in localized prostate cancer intensity-modulated radiation therapy (IMRT).
- Locating the prostatic urethra in planning computed tomography (pCT) images presents a significant challenge.
Purpose of the Study:
- To develop and assess the feasibility of a multiatlas-based auto-segmentation method for prostatic urethra identification.
- The method utilizes deformable image registration (DIR) accuracy prediction with machine learning (ML).
Main Methods:
- A novel method was developed using a machine learning (ML) model (support vector machine regression - SVR) to predict deformable image registration (DIR) accuracy for selecting optimal atlases.
- Structure-based DIR was employed to deform selected atlases and propagate urethra contours.
- The method was trained and validated on 120 prostate cancer patients, with accuracy evaluated using centerline distance (CLD) against manual contouring (ground truth).
Main Results:
- The proposed method achieved the highest accuracy, with a mean CLD of 2.09 ± 0.89 mm, significantly outperforming existing methods (Acosta et al.: 2.77 ± 0.99 mm, P = 0.022; Waterman et al.: 3.47 ± 1.19 mm, P < 0.001).
- Accuracy was particularly improved in the upper third of the prostatic urethra.
- The method demonstrated a mean error of 2.09 mm in identifying the prostatic urethra.
Conclusions:
- A novel multiatlas-based auto-segmentation method incorporating DIR accuracy prediction with ML was successfully developed for prostatic urethra identification.
- This method offers high accuracy and has the potential to replace the use of temporary indwelling urinary catheters in prostate cancer IMRT.
- The combined effects of SVR model patient selection, modified atlas characteristics, and the DIR algorithm contribute to the method's effectiveness.
