Semi-automatic segmentation of subcutaneous tumours from micro-computed tomography images

Area of Science:

• Biomedical Imaging
• Medical Image Analysis
• Cancer Research Technology

Background:

• Preclinical subcutaneous tumors are vital in cancer research.
• Micro-computed tomography (microCT) offers limited tissue contrast, complicating tumor boundary identification.
• Accurate tumor segmentation is crucial for quantitative analysis in preclinical studies.

Purpose of the Study:

• To develop and validate a semi-automatic algorithm for segmenting subcutaneous tumors from microCT image data.
• To address the challenge of low tissue contrast in microCT for precise tumor boundary detection.
• To establish a foundation for advanced preclinical cancer research using open-source tools.

Main Methods:

• Utilized local phase feature detection to identify faint boundary features in microCT images.
• Employed a level set-based active contour model for smooth contour generation along sparse boundary data.
• Validated the segmentation algorithm against manual expert delineations and micro-positron emission tomography (microPET) data.

Main Results:

• The algorithm consistently segmented over 70% of the tumor region across various cases and volumes when compared to manual segmentations.
• The segmentation captured more than 80% of the functional tumor volume as indicated by microPET data.
• Demonstrated the feasibility of segmenting tumors from microCT images without exogenous contrast agents.

Conclusions:

• The developed semi-automatic segmentation method is effective for preclinical subcutaneous tumors in microCT.
• This approach provides a reliable and quantitative tool for cancer research, enhancing the utility of microCT data.
• The open-source availability of the code facilitates further research and application in the field.