Abstract 1251P
Background
Recent advances in medical imaging technology allow the use of advanced image analysis methods by the extraction of quantitative features from medical images to characterize tumor biology in order to provide informations that may be useful to guide therapies and predict survival. In our previous experience, a radiomics signature mixing semantic and image-based features was able to predict the reduction of the target volume during chemoradiation in LA-NSCLC. The aim of this study is to investigate which 3D ROI contain more quantitative information that can be exploit for a radiomic approach in lung cancer.
Methods
We studied 122 LA- NSCLC treated with chemoradiation with or without adaptive approach, in order to investigate the predictive power of the radiomic features on tumor shrinkage. Gross Tumour Volume (GTV), Clinical Tumour Volume (CTV) and Planned Tumour Volume (PTV) were manually delineated by expert radiation oncologists, providing three 3D ROIs. For each ROI we computed 105 quantitative biomarkers using the opensource software platform PyRadiomics. The features belong to the following subgroups: Shape based measures, First Order Statistics, Gray Level Cooccurence Matrix, Gray Level Run Length Matrix, Gray Level Size Zone Matrix, Neighbouring Gray Tone Difference Matrix and Gray Level Dependence Matrix. For feature selection and classification ee use leave-one-patient out approach and, for each run, we proceeded ranked all the features using the recursive feature elimination (RFE) method and then we trained an AdaBoost meta-classifier. Such a training-testing procedure was repeated adding every time a new feature according to the RFE output.
Results
The largest accuracy (80.3%) is obtained using 16 features computed from CTV ROIs. From the 16 features which gave the best result 5 belonged to the shape feature class, 5 to the First oder Statistics, 2 to the class GLDM, 2 to the NGTDM class, and one feature belonged to the classes GLCM and GLSZM, respectively.
Conclusions
These preliminary results suggest that there may also be useful information in healthy tissue bordering on neoplastic disease macroscopically evident in imaging. Further work is needed to substantiate the proposed hypothesis.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.