Perioperative Chemotherapy with FLOT Scheme in Resectable Gastric Adenocarcinoma: A Preliminary Correlation between TRG and Radiomics

• Published in: Applied sciences Vol. 11; no. 19; p. 9211
• Main Authors: Garbarino, Giovanni Maria, Zerunian, Marta, Berardi, Eva, Mainardi, Federico, Pilozzi, Emanuela, Polici, Michela, Guido, Gisella, Rucci, Carlotta, Polidori, Tiziano, Tarallo, Mariarita, Laracca, Giovanni Guglielmo, Iannicelli, Elsa, Mercantini, Paolo, Annibale, Bruno, Laghi, Andrea, Caruso, Damiano
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.10.2021
• Subjects: 5-Fluorouracil; Adenocarcinoma; Biomarkers; Chemotherapy; Gastric cancer; Gastrointestinal surgery; Medical imaging; Medical prognosis; Multidisciplinary teams; Oxaliplatin; Patients; perioperative chemotherapy; Radiomics; response to treatment; Tumors; Ultrasonic imaging; United States > US; Europe
• ISSN: 2076-3417; 2076-3417
• DOI: 10.3390/app11199211

Perioperative chemotherapy (p-ChT) with a fluorouracil plus leucovorin, oxaliplatin, and docetaxel (FLOT) scheme is the gold standard of care for locally advanced gastric cancer. We aimed to test CT radiomics performance in early response prediction for p-ChT. Patients with advanced gastric cancer who underwent contrast enhanced CT prior to and post p-ChT were retrospectively enrolled. Histologic evaluation of resected specimens was used as the reference standard, and patients were divided into responders (TRG 1a-1b) and non-responders (TRG 2-3) according to their Becker tumor regression grade (TRG). A volumetric region of interest including the whole tumor tissue was drawn from a CT portal-venous phase before and after p-ChT; 120 radiomic features, both first and second order, were extracted. CT radiomics performances were derived from baseline CT radiomics alone and ΔRadiomics to predict response to p-ChT according to the TRG and tested using a receiver operating characteristic (ROC) curve. The final population comprised 15 patients, 6 (40%) responders and 9 (60%) non-responders. Among pre-treatment CT radiomics parameters, Shape, GLCM, First order, and NGTDM features showed a significant ability to discriminate between responders and non-responders (p < 0.011), with Cluster Shade and Autocorrelation (GLCM features) having AUC = 0.907. ΔRadiomics showed significant differences for Shape, GLRLM, GLSZM, and NGTDM features (p < 0.007). MeshVolume (Shape feature) and LongRunEmphasis (GLRLM feature) had AUC = 0.889. In conclusion, CT radiomics may represent an important supportive approach for the radiologic evaluation of advanced gastric cancer patients.