Enhancing PET/CT Radiomics Robustness Through Graph Signal Processing

Area of Science:

• Medical Imaging and Radiomics
• Oncology and Cancer Diagnostics
• Graph Signal Processing in Medicine

Background:

• Prostate cancer bone metastases cause significant complications, necessitating accurate imaging.
• Classical radiomics faces limitations in robustness and capturing lesion heterogeneity.
• Need for advanced imaging analysis for reliable diagnosis and treatment planning.

Purpose of the Study:

• Introduce a translational graph-based radiomics approach for improved bone lesion characterization.
• Develop novel quantitative descriptors for enhanced robustness and clinical reliability.
• Overcome limitations of classical radiomics in handling inter-scanner variability and segmentation dependence.

Main Methods:

• Derived graph representation from PET/CT bone lesions using point clouds and Delaunay triangulation.
• Extracted orientation, connectivity, and transform-based graph features using signal processing.
• Evaluated feature robustness, redundancy, and classification performance (LDA, SVM) against classical radiomics.

Main Results:

• Graph-based features captured non-redundant information and showed superior robustness to batch effects and segmentation variability.
• Classification models using proposed features outperformed classical radiomics and combined sets.
• Achieved mean AUC of 72.14% (LDA) and 66.49% (SVM) with proposed features.

Conclusions:

• The graph-based radiomics framework offers improved robustness and diagnostic performance for PET/CT bone lesions.
• This approach shows promise as an integrative tool for reliable prostate cancer bone metastasis characterization.
• Potential to enhance clinical decision-making in managing prostate cancer patients with bone lesions.