LAUSR

Purpose Radiomics have proved to be useful in exploiting imaging data to correlate and even predict some phenotypic or genetic tumor characteristics. Our purpose was to build and test a complete radiomic analysis process, mainly addressed to CT imaging of thoracic district, using a 3D-Slicer platform (mainly promoted by NIH) and R computing and graphic software (international R-core team). Methods and materials 3D-Slicer free software (v.4.6 and 4.7) was installed on a 32 GB-RAM workstation, equipped with two diagnostic quality 3MP monitors. Lesions’ outlining exploited the semi-automatic 3D-Slicer contouring tool Fast-GrowCut. Plastimatch and Radiomics modules were used, respectively, for images elastic registration (among CT scans with and without contrast agent) and for 94 radiomic features determination (first order, shape, GLCM, GLRLM, GLSZM). The full process was tested on 21 patients, all affected by NSCLC, scanned by the same tomograph. Features were assessed only on non-enhanced images. An in-house R script was built to evaluate the main statistical indices and some graphical results on the test group. Results 3D-Slicer free platform has proved an excellent one for accurate lesions’ delineation on thorax CT images, thanks to the Fast-GrowCut tool which has high performances in the presence of high HU contrast. The elastic fusion transformation matrix (Plastimatch) helped to better define the tumor contour, excluding atelectasis or border regions in some unclear cases, if the correspondent enhanced scan was available. Radiomics tool (3D-Slicer option by Pyradiomics), combined with the in-house R routine, permitted to evaluate intensive and exstensive radiomic features, together with their densities when meaningful (e.g. entropy), to evaluate their distribution and prepare for radiomic fingerprint maps and cluster analysis. Conclusion Our works has built a free of charge complete radiomics process, based on internationally validated softwares, to contour lesions and assess radiomic features, features densities and statistical analysis on patients’ groups.