Related Experiment Videos
Marker-controlled watershed for lymphoma segmentation in sequential CT images.
Jiayong Yan1, Binsheng Zhao, Liang Wang
1Medical Physics Department, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, New York 10021, USA. yanj12@mskcc.org
Medical Physics
|August 11, 2006
Summary
A new semi-automated algorithm reliably segments lymphoma in CT scans using marker-controlled watershed transform. This method accurately quantifies tumor volume, aiding in diagnosis and treatment planning for lymphoma patients.
Area of Science:
- Medical Imaging
- Computational Pathology
- Oncology
Background:
- Lymphoma segmentation in CT images is challenging due to variations in tumor location, intensity, and tissue contrast.
- Accurate segmentation is crucial for diagnosis, treatment planning, and monitoring lymphoma progression.
Purpose of the Study:
- To develop and evaluate a reliable, semi-automated marker-controlled watershed algorithm for segmenting lymphoma in sequential CT images.
- To assess the algorithm's accuracy in quantifying lymphoma volume and its boundaries.
Main Methods:
- A novel marker-controlled watershed algorithm was developed, utilizing automatic internal marker determination (Canny edge detection, thresholding, morphological operations, distance map) and slice-to-slice propagation for external markers.
- The algorithm was applied to 29 lymphomas in nine patients, with results compared against manual delineations by a blinded radiologist.
Main Results:
- The algorithm achieved a mean overlap ratio of 83.2%, with 13.5% overestimation and 5.5% underestimation.
- Quantitative evaluation showed a mean average boundary distance of 0.7 mm and a Hausdorff boundary distance of 3.7 mm.
Conclusions:
- The proposed semi-automated algorithm demonstrates potential for reliable segmentation and quantification of lymphomas in CT images.
- This approach offers a practical tool for improving diagnostic accuracy and treatment assessment in lymphoma management.