Radiomics analysis of 3D dose distributions to predict toxicity of radiotherapy for lung cancer

Purpose (Chemo)–radiotherapy (RT) is the gold standard treatment for patients with locally advanced lung cancer non accessible surgery. However, current toxicity prediction models rely on clinical and dose volume histograms (DVHs) remain unsufficient. The goal of this work to investigate added predictive value radiomics approach applied maps regarding acute late toxicities in both lungs esophagus. Methods Acute scored using CTCAE v4.0 were retrospectively collected treated RT our institution. Radiomic features extracted from 3D considering Gy values as grey-levels images. DVH usual factors also considered. Three (clinical only, + combined, i.e., including radiomics) incrementally trained a neural network 70% grade ≥2 pulmonary (APT/LPT) esophageal (AET). After bootstrapping (n = 1000), optimal cut-off determined based Youden Index. then evaluated remaining 30% balanced accuracy (BAcc). Results 167 2015 2018: 78% small-cell cancers, 14% cancers 8% other histology median age at 66 years. Respectively, 22.2%, 16.8% 30.0% experienced APT, LPT AET. In training set 117), corresponding BAcc only/clinical DVH/combined 0.68/0.79/0.92, 0.66/0.77/0.87 0.68/0.73/0.84. testing evaluation 50), these obtained 0.69/0.69/0.92, 0.76/0.80/0.89 0.58/0.73/0.72. Conclusion RT, radiomic seem surpass DVHs APT LPT.