Related Experiment Video
Updated: Sep 30, 2025
05:19
Magnetic Resonance Imaging Assessment of Carcinogen-induced Murine Bladder Tumors
Published on: March 29, 2019
10.5K
PIxel-Level Segmentation of Bladder Tumors on MR Images Using a Random Forest Classifier
Ziqi Li1, Na Feng2, Huangsheng Pu3
1School of Biomedical Engineering, 12644Air Force Medical University, Xi'an, PR China.
Technology in Cancer Research & Treatment
|March 17, 2022
Summary
A novel pixel-level random forest method accurately segments bladder cancer (BCa) on MR images. This approach achieves high performance, improving tumor identification for better clinical management of BCa patients.
Area of Science:
- Medical Imaging
- Oncology
- Machine Learning
Background:
- Bladder cancer (BCa) detection relies on MR imaging, where wall thickening indicates malignancy.
- Accurate tumor segmentation is crucial for staging and grading BCa, guiding clinical decisions.
Purpose of the Study:
- To develop and evaluate a new pixel-level feature-based random forest (RF) method for precise BCa segmentation on MR images.
- To improve the accuracy of noninvasive pathological staging and grading of BCa.
Main Methods:
- Utilized high-throughput pixel-level features and a random forest classifier for BCa segmentation.
- Trained the model using regions of interest (ROIs) including tumor and bladder wall.
- Segmented candidate regions containing BCa and adjacent wall tissue in the testing set.
Main Results:
- Evaluated on 56 BCa patients, the method demonstrated a mean Dice Similarity Coefficient (DSC) of 0.906.
- Achieved a mean Average Symmetric Surface Distance (ASSD) of 1.190 mm.
- Outperformed existing state-of-the-art methods in tumor region separation.
Conclusions:
- The proposed pixel-level BCa segmentation method provides accurate lesion identification on MR images.
- This approach shows significant potential for enhancing the noninvasive assessment of bladder cancer.
