Related Experiment Video
Updated: Aug 8, 2026
08:39
Longitudinal Micro-Computed Tomography Image Analysis for User-Defined Region of Interest in Critical-Sized Bone Defects
Published on: June 24, 2025
Model-based analysis of local shape for lesion detection in CT scans
Paulo R S Mendonça1, Rahul Bhotika, Saad A Sirohey
1GE Global Research, One Research Circle, Niskayuna, NY 12309, USA. mendonca@research.ge.com
Medical Image Computing and Computer-Assisted Intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention
|May 12, 2006
Summary
This study introduces a novel computer tomography (CT) analysis technique for automated pulmonary lesion detection. The method combines geometric and intensity models with local curvature analysis to improve early lung cancer diagnosis.
Area of Science:
- Medical Imaging
- Radiology
- Computer-Aided Diagnosis
Background:
- Thin-slice computed tomography (CT) offers high-resolution imaging crucial for early lung cancer detection.
- Large CT data volumes can lead to variability in radiological interpretations, necessitating automated detection systems.
- Accurate identification of pulmonary lesions is vital for timely and effective cancer treatment.
Purpose of the Study:
- To develop and evaluate an automated technique for detecting pulmonary lesions in CT scans.
- To combine geometric and intensity models with local curvature analysis for enhanced lesion identification.
- To assess the algorithm's performance against expert-determined ground truth and its potential to improve radiologist sensitivity.
Main Methods:
- A novel technique integrating geometric and intensity models with local curvature analysis was developed.
- Local shape at each voxel is represented using principal curvatures of its isosurface without explicit extraction.
- Voxel classification into anatomical structures (nodules, vessels) is achieved by comparing curvatures to analytical shape models.
Main Results:
- The algorithm was evaluated on 242 CT exams with expert-determined ground truth.
- Performance was quantified using free-response receiver-operator characteristic curves.
- The study demonstrated the algorithm's potential to improve radiologist sensitivity in detecting pulmonary lesions.
Conclusions:
- The proposed technique offers a promising approach for automated pulmonary lesion detection in CT imaging.
- Combining shape and intensity analysis with local curvature provides a robust method for anatomical structure identification.
- This automated system has the potential to aid radiologists in the early diagnosis of lung cancer, improving patient outcomes.
