A computerized volumetric segmentation method applicable to multi-centre MRI data to support computer-aided breast
Gokhan Ertas1,2, Simon J Doran3, Martin O Leach1
1Cancer Research UK Cancer Imaging Centre, Division of Radiotherapy and Imaging, The Institute of Cancer Research, 123 Old Brompton Road, London, SW7 3RP, UK.
Medical & Biological Engineering & Computing
|April 24, 2016
Summary
A new computerized algorithm accurately segments breast tissue in MRI scans. This fast, automated method is suitable for multi-centre studies, improving density assessment and lesion localization.
Area of Science:
- Medical Imaging
- Radiology
- Computer-Aided Diagnosis
Background:
- Accurate breast MRI segmentation is crucial for density assessment and lesion detection.
- Existing methods may lack efficiency or suitability for multi-centre data.
Purpose of the Study:
- To develop a fast, computerized algorithm for volumetric breast segmentation applicable to multi-centre MRI data.
- To automate breast segmentation without requiring prior anatomical information.
Main Methods:
- Employed 3D bias-corrected fuzzy c-means clustering and morphological operations.
- Determined full breast extent on T1-weighted images.
- Utilized automatic midsternum detection for left and right breast identification.
Main Results:
- The algorithm demonstrated high performance in a UK multi-centre study involving 82 women.
- Achieved excellent agreement with manual segmentation: Relative Overlap (RO) 0.94 ± 0.05, True-Positive Volume Fraction (TPVF) 0.97 ± 0.03, False-Positive Volume Fraction (FPVF) 0.04 ± 0.06.
- Consistent performance observed on both training and test datasets.
Conclusions:
- The developed algorithm provides a fast and accurate solution for volumetric breast segmentation in multi-centre MRI screening.
- This automated approach enhances the reliability of breast density assessment and lesion localization.
