Presentation: An international standard compliant tool for radiomic feature extraction from medical images
Date
2022-06-02
Authors
Summary
Cancer is a heterogeneous disease with unique genomic and phenotypic characteristics that differ from patient to patient and even from tumor region to tumor region. Precision oncology is therefore an important field in cancer research as it allows for specific treatments for each cancer patient based on their genomic profile. This genomic heterogeneity of aggressive tumors translates into heterogeneous tumor characteristics at the anatomical scale that can be captured using quantitative image analysis. The latter will give an ability to make accurate tumor phenotypic predictions. The extraction of these features is known as “radiomic analysis” which is performed using several mathematical algorithms to extract these features. A platform under the name MEDomicsLab which is a result of an international collaboration (MEDomics Consortium) has been developed for the extraction of radiomic features from original or previously filtered images in order to highlight discriminating features in these medical images. The validation of these image analyses remains a major challenge. Therefore, an international collaboration under the name IBSI (Image Biomarker Standardization Initiative) seeks to standardize the extraction of these image features through the implementation of several standards and guidelines in order to establish a consensus between the different participating teams/universities. The standardization has been separated into two chapters, IBSI 1 and IBSI 2. IBSI 1 focuses on the standardization of radiomic feature calculations. On the other hand, IBSI 2, which is currently under development, targets the standardization of the implementation of convolutional filters (e.g., Gaussian Laplacian, wavelet-based filter) as well as the extraction of radiomic features from the previously filtered images. We have integrated the procedures established by IBSI into the MEDomicsLab platform. Our results show consensus with the majority of IBSI collaborators for both chapters. The MEDomicsLab tool will be made available free of charge to researchers in Quebec. It is expected that this tool for extracting rich and complex radiomic features will be used to conduct multicenter studies in precision oncology.