ESTRO 2021 Abstract Book

S1548

ESTRO 2021

ACC v =0.8) were obtained respectively for training and validation sets. Figure 1 shows the roc curve obtained for the MLM (TR_GAUSS, SVM, elastic-net linear classifier).

Conclusion In this paper, we demonstrated that MLM models trained on the radiomics, extracted from DWI and corresponding ADC maps, could potentially predict the grade of CC tumors. Future work will consist on consolidating our results on more patient datasets. PO-1819 Prediction of rectal cancer tumor response with MRI-based clinical Radiomics Model C. Rosa 1,2 , F.C. Di Guglielmo 2 , L. Gasparini 2 , L. Caravatta 2 , M. Di Tommaso 2 , A. Delli Pizzi 1 , M. D'Annibale 1 , P. Chiacchiaretta 3 , A.M. Chiarelli 3 , P. Croce 3 , D. Genovesi 2,1 1 G. D’Annunzio University, Department of Neuroscience, Imaging and Clinical Sciences, Chieti, Italy; 2 SS. Annunziata Hospital, Department of Radiation Oncology, Chieti, Italy; 3 G. D’Annunzio University, ITAB-Institute of Advanced Biomedical Technologies, Chieti, Italy Purpose or Objective Neoadjuvant chemoradiotherapy (CRT) followed by total mesorectal excision (TME) is a standard treatment for locally advanced rectal cancer (LARC) patients. Magnetic Resonance Imaging (MRI) is the gold standard both for local staging and treatment response assessment. New methods extracting data from clinical images, based on automatic extraction of features from radiological images (radiomic features) and machine learning approaches, revealed promising results combined to morphologic and clinical assessment. Therefore, a novel approach is to investigate pre-treatment MRI-based biomarkers for treatment response prediction. We present a novel machine learning model combining pre-treatment MRI- based clinical and radiomic features for early prediction of treatment response in LARC patients.