Related Experiment Video
Updated: Mar 16, 2026
Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
Generic and robust method for automatic segmentation of PET images using an active contour model.
Mingzan Zhuang1, Rudi A J O Dierckx2, Habib Zaidi3
1Department of Nuclear Medicine and Molecular Imaging, University of Groningen, University Medical Center Groningen, 9700 RB Groningen, The Netherlands; Department of Radiation Oncology, Tumor Hospital of Shantou University Medical College, Shantou, Guangdong 515000, China; and The Key Laboratory of Digital Signal and Image Processing of Guangdong Province, Shantou University, Shantou, Guangdong 515000, China.
This study introduces MASAC, an automatic method for delineating tumor volumes in PET images, improving accuracy for radiation therapy. MASAC demonstrates robust performance in phantom and clinical studies, outperforming other methods in key segmentation metrics.
Area of Science:
- Medical Imaging
- Radiotherapy
- Computational Biology
Background:
- Positron emission tomography (PET) imaging offers potential for enhancing radiation therapy planning and treatment response assessment.
- Distinguishing tumor boundaries from normal tissue in PET images is challenging due to low spatial resolution and noise.
- Accurate tumor volume delineation is crucial for effective radiation therapy and outcome evaluation.
Purpose of the Study:
- To develop a generic and robust method for automatic tumor volume delineation using an active contour model.
- To evaluate the performance of the developed method using both simulated phantom and clinical studies.
- To compare the proposed method against existing segmentation techniques.
Main Methods:
- Developed MASAC (Method for Automatic Segmentation using Active Contour model), integrating histogram fuzzy C-means clustering and localized textural information.
- Employed the lattice Boltzmann method for efficient solving of the level set equation, enabling faster computation and parallel programming.
- Evaluated performance on 20 simulated phantom studies and 16 clinical studies (pharyngolaryngeal squamous cell carcinoma and nonsmall cell lung cancer), comparing with contourlet-based active contour (CAC) and Schaefer's thresholding (ST) methods.
Main Results:
- MASAC and CAC showed similar segmentations in phantom studies, while ST yielded unreliable results.
- In clinical studies, MASAC achieved the best Dice Similarity Coefficient (DSC) of 0.71 ± 0.09 and Classification Error (CE) of 53.92% ± 12.65%.
- MASAC demonstrated reliable quantification across lesion types with good accuracy, achieving mean RE of -13.35% ± 11.87% (PSs) and -11.15% ± 23.66% (CSs), mean DSC of 0.89 ± 0.05 (PSs) and 0.71 ± 0.09 (CSs), and mean CE of 19.19% ± 7.89% (PSs) and 53.92% ± 12.65% (CSs).
Conclusions:
- The novel PET segmentation algorithm (MASAC) is applicable to various lesion types, providing accurate and consistent target volume delineations.
- The method has the potential to reduce intraobserver and interobserver variabilities associated with manual delineation.
- Improved accuracy in treatment planning and outcome evaluation is anticipated with the use of this automated segmentation technique.
