Related Experiment Video
Updated: Feb 27, 2026
09:36
Outer-Boundary Assisted Segmentation and Quantification of Trabecular Bones by an Imagej Plugin
Published on: March 14, 2018
9.8K
Automatic MPST-cut for segmentation of carpal bones from MR volumes
Laura Gemme1, Sonia Nardotto1, Silvana G Dellepiane1
1Department of Electrical, Electronic, Telecommunications Eng. and Naval Architecture (DITEN), Via Opera Pia 11a, Genova, 16145, Università degli Studi di Genova, Italy.
Computers in Biology and Medicine
|June 27, 2017
Summary
This study introduces a novel semi-automatic method for segmenting carpal bones in MRI scans, crucial for diagnosing rheumatoid arthritis (RA). The technique accurately identifies bone structures, improving diagnostic capabilities for rheumatic diseases.
Area of Science:
- Medical Imaging
- Rheumatology
- Computer Vision
Background:
- Magnetic Resonance Imaging (MRI) offers higher sensitivity for detecting Rheumatoid Arthritis (RA) signs compared to conventional radiology.
- Accurate carpal bone segmentation is vital for quantitative diagnosis, erosion assessment, and multi-temporal data analysis in RA.
- Existing segmentation methods may lack adaptability or require extensive prior knowledge.
Purpose of the Study:
- To propose a new, semi-automatic, 3D graph-based segmentation method for carpal bone extraction from MRI data.
- To develop an unsupervised and adaptive approach that does not rely on a priori models.
- To evaluate the method's efficiency and accuracy for bone segmentation in T1-weighted MR volumes.
Main Methods:
- A semi-automatic, 3D graph-based segmentation approach utilizing a Minimum Path Spanning Tree (MPST) and a novel MPST-cut criterion based on compactness shape factor.
- The process involves unsupervised, adaptive stages: cost-labeling and graph-cutting, initiated by selecting a source point within the Region of Interest (ROI).
- Applied to a database of 96 T1-weighted MR bone volumes, with performance validated against manual segmentations by rheumatologists.
Main Results:
- The proposed method demonstrates efficient and satisfactory performance for carpal bone segmentation on low-field MR volumes.
- Quantitative evaluation using metrics from the confusion matrix shows comparable or improved results against existing literature.
- The technique successfully extracts carpal bone data, adapting to individual anatomical variations without prior models.
Conclusions:
- The developed semi-automatic, 3D graph-based segmentation method provides an effective tool for carpal bone analysis in the context of rheumatic diseases.
- This unsupervised and adaptive approach enhances the potential of MRI in Rheumatoid Arthritis diagnosis and monitoring.
- The method's satisfactory performance on low-field MR volumes suggests its clinical utility and broad applicability.
Keywords:
Automatic volume segmentationCarpal bonesGraph-segmentationMPST-cutMagnetic Resonance volumesShape factor
