SAFARI: shape analysis for AI-segmented images

Area of Science:

• Medical imaging analysis
• Computational pathology
• Radiomics

Background:

• Medical image analysis often involves segmenting regions of interest (ROI).
• Current ROI analysis methods are inconsistent and vary significantly between studies.
• A standardized tool is needed to convert ROIs into analyzable shape representations and features.

Purpose of the Study:

• To develop an open-source R package and online toolkit for shape analysis of segmented medical images.
• To provide a user-friendly platform for ROI labeling and shape feature extraction.

Main Methods:

• Developed SAFARI (shape analysis for AI-segmented images), an R package and online toolkit.
• Extracted shape features from segmented maps generated by AI or manual segmentation.
• Applied SAFARI to case studies involving lung cancer and glioblastoma patients.

Main Results:

• SAFARI facilitates efficient and user-friendly segmentation and analysis of ROIs in medical images.
• Half of the shape features extracted by SAFARI demonstrated significant associations with survival outcomes.
• Demonstrated utility in case studies of 143 lung cancer patients and 61 glioblastoma patients.

Conclusions:

• SAFARI is an efficient, easy-to-use toolkit for medical image ROI analysis.
• The package is available on CRAN and via an online portal.
• SAFARI aids in extracting clinically relevant shape features for survival outcome prediction.