Pulmonary tumor volume delineation in PET images using deformable models

Area of Science:

• Medical Imaging
• Oncology
• Computational Biology

Background:

• Lung cancer is a leading cause of cancer-related mortality globally.
• Accurate tumor segmentation in PET data is crucial for effective treatment planning.
• Manual segmentation is time-consuming, prone to errors, and causes physician fatigue due to the dynamic nature of lung tumors during respiration.

Purpose of the Study:

• To develop and validate a semi-automatic method for segmenting and tracking lung tumors in PET data.
• To overcome the limitations of manual segmentation, including errors and inter-observer variability.
• To provide a consistent and efficient tool for oncologists in lung cancer management.

Main Methods:

• Initial tumor segmentation in the first PET frame using wavelet features and a Support Vector Machine (SVM) classifier.
• Utilizing the expert-validated segmented tumor from the first frame to initialize contour segmentation in subsequent frames via the level set method.
• Implementing the tumor volume from the first frame as a termination criterion for the level set method to ensure consistent segmentation.

Main Results:

• The proposed semi-automatic method successfully segments and tracks lung tumors in PET data.
• The technique eliminates the need for manual tumor segmentation, significantly reducing physician workload and potential errors.
• Segmentation results are consistent and operator-independent, unlike traditional manual methods.

Conclusions:

• The developed semi-automatic method offers a promising alternative to manual lung tumor segmentation in PET imaging.
• This approach enhances accuracy, consistency, and efficiency in lung cancer treatment planning.
• The operator-independent nature of the technique ensures reliable and reproducible tumor volume assessment.