305. Can magnetic-resonance-imaging volumetric texture analysis predict treatment outcome in rectal cancer patients undergoing neoadjuvant chemo-radiation?

Purpose This work aims to evaluate the possibility to use volumetric Texture analysis (TA) of T2 images and ADC maps to predict the clinical output in patients affected by advanced rectal cancer (LARC). Methods For this work were considered patients with non-mucinous extraperitoneal rectum adenocarcinoma with different degree of differentiation treated with neoadjuvant chemo-radiotherapy (C-RT) and radical surgery (TME). The Mandard score permitted to establish the tumour regression grade (TRG). Enrolled patients were treated with C-RT (45 Gy + 9 Gy 1,8 Gy/day and oxaliplatin and capecitabine) and subjected to pelvic MRI (T2 and DWI) examinations (1.5-T system, Signa Excite HD, GE) made at baseline and 40  ± 20 days after the end of C-RT. Two blinded readers manually segmented tumour volumes and used a homemade Image macro to evaluate TA parameters. The application of a Laplacian of Gaussian bandpass filter between 0.5 and 2.5 permits to highlight different spatial scales. Using SPSS software Pearson correlation, Cox regression, ROC and Kaplan Meier curve were evaluated considering TA parameters (TAp), TRG and metastasis development. Were correlated with early progression of disease (ePD) omitting the variable with the lowest Pearson correlation coefficient to avoid a model overfitting. With k-fold and ROC analysis was possible to validate our performance model. Results This study included 23 patients (average age 65,9 ± 11,7 years): 9 were TRG1 and 14 TRG2. No statistically significant correlation was found considering TAp calculated from T2 images. From ADC images emerged a significant correlation between tumour stage (p = 0.016) and metastasis development (p = 0.013). Moreover, ROC analysis showed how this parameter is predictive for complete treatment response (specificity 86% and accuracy 83%). Conclusion This work shows how TAp could predict the outcome of patients to implement a personalised treatment.