Abstract
Radiomics is a quantitative approach to medical imaging consisting in the extraction of a large number of features (fts) from diagnostic images that can be associated with tumour pathophysiology and converted into mineable high-dimensional data. We define a model of workflow for extraction and selection of radiomics fts from the baseline computed tomography scan images of stage III NSCLC pts, thus creating a radiomics profile to guide the clinical decision-making process. We retrospectively collected data of stage III NSCLC pts referred to Veneto Institute of Oncology from 2012 to 2021. The radiomics pipeline includes (1) the definition of inclusion/exclusion criteria based on image quality and clinico-pathological data, (2) data selection for training and validation cohorts, (3) image segmentation and annotation of target lesions, (4) fts extraction, (5) fts reduction and selection, (6) radiomics signature building. Image segmentation and fts extraction were performed with a commercial software (HealthMyne® Platform), radiomics analysis using the open-source package RadAR (Radiomics Analysis with R) for the statistical language R. On a training cohort of 60 stage III NSCLC pts, we performed the volume segmentation of a maximum of 4 target lesions for each patient. 517 fts were extracted and categorized into morphology-, intensity-, texture- and filter-based fts. A first stage univariate analysis of the primitive lung target lesion was performed. The features selection was made combining different approaches: a principal component analysis to reduce fts redundancy, an outlier analysis to exclude fts whose distribution showed far outliers, and a further selection on the basis of concordance with clinical data performed with the criterion of minimum redundancy maximum relevance. The correlation of the 10 fts resulting from this pipeline with each relevant clinical data can then be quantitatively evaluated through Spearman's non-parametric coefficient. A radiomic workflow for Stage III NSCL data has been studied. A validation cohort study is warranted in order to combine tumor phenotype characteristics through artificial intelligence applications into predictive and prognostic models.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.