Related Experiment Video
Updated: Jul 9, 2026
08:40
Quantitation of Protein Expression and Co-localization Using Multiplexed Immuno-histochemical Staining and Multispectral Imaging
Published on: April 8, 2016
12.8K
Interactive Cascaded Network for Prostate Cancer Segmentation from Multimodality MRI with Automated Quality
Weixuan Kou1, Cristian Rey2, Harry Marshall3
1Department of Electrical Engineering, City University of Hong Kong, Hong Kong.
Bioengineering (Basel, Switzerland)
|August 29, 2024
Summary
This study introduces a new framework for prostate cancer (PCa) segmentation using AI. It significantly reduces manual annotation, achieving high accuracy comparable to full manual segmentation with only half the effort.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Accurate prostate cancer (PCa) segmentation from multiparametric MRI is vital for clinical decisions.
- Current automated methods lack accuracy, while interactive methods are time-consuming.
- Existing segmentation workflows face challenges in cost-effectiveness and user burden.
Purpose of the Study:
- To develop an innovative framework for accurate and efficient PCa segmentation.
- To reduce the manual annotation burden in medical image segmentation.
- To improve the cost-effectiveness of segmentation workflows while maintaining high accuracy.
Main Methods:
- A novel framework combining a coarse segmentation network, a rejection network, and the Segment Anything Model (SAM).
- Automated initial segmentation followed by quality assessment via a rejection network.
- Selective user interaction for low-quality segmentations and automated ROI cropping for high-quality ones.
Main Results:
- The framework significantly reduces annotation effort by flagging ~20% of images for manual review.
- Achieved final segmentation accuracy statistically indistinguishable from full manual annotation with only 50% manual input.
- Demonstrated substantial improvements in segmentation efficiency and accuracy.
Conclusions:
- The proposed framework effectively balances segmentation accuracy and efficiency.
- It offers a cost-effective solution for PCa segmentation, reducing user intervention.
- The adaptable framework holds potential for various medical image segmentation applications.
