Fractal-Based Radiomic Approach to Tailor the Chemotherapy Treatment in Rectal Cancer: A Generating Hypothesis Study

IntroductionThe aim of this study was to create a radiomic model able to calculate the probability of 5-year disease-free survival (5yDFS) when oxaliplatin (OXA) is or not administered in patients with locally advanced rectal cancer (LARC) and treated with neoadjuvant chemoradiotherapy (nCRT), allowing physicians to choose the best chemotherapy (CT) regimen. MethodsLARC patients with cT3-4 cN0 or cT1-4 cN1-2 were treated according to an nCRT protocol that included concomitant CT schedules with or without OXA and radiotherapy dose of 55 Gy in 25 fractions. Radiomic analysis was performed on the T2-weighted (T2-w) MR images acquired during the initial tumor staging. Statistical analysis was performed separately for the cohort of patients treated with and without OXA. The ability of every single radiomic feature in predicting 5yDFS as a univariate analysis was assessed using the Wilcoxon-Mann-Whitney (WMW) test or t-test. Two logistic models (one for each cohort) were calculated, and their performance was assessed using the area under the receiver operating characteristic (ROC) curve (AUC). ResultsA total of 176 image features belonging to four families (morphological, statistical, textural, and fractal) were calculated for each patient. At the univariate analysis, the only feature showing significance in predicting 5yDFS was the maximum fractal dimension of the subpopulation identified considering 30% and 50% as threshold levels (maxFD(30-50)). Once the models were developed using this feature, an AUC of 0.67 (0.57-0.77) and 0.75 (0.56-0.95) was obtained for patients treated with and without OXA, respectively. A maxFD(30-50) >1.6 was correlated to a higher 5yDFS probability in patients treated with OXA. ConclusionThis study suggests that radiomic analysis of MR T2-w images can be used to define the optimal concomitant CT regimen for stage III LARC cancer patients. In particular, by providing an indication of the gross tumor volume (GTV) spatial heterogeneity at ini