Medical Oncology Resident University of Minnesota Saint Paul, Minnesota
Tumor heterogeneity is a well-established marker of tumor behavior and it has been associated with prognosis in human lung tumors. Quantitative analysis of computed tomography (CT) radiomic features is an indirect measure of tumor heterogeneity. The purpose of this study was to extract CT radiomic features from canine primary pulmonary tumors and correlate features to histopathologic diagnosis or survival. First-order statistical-based CT texture features were extracted from segmented tumor volumes. Time to tumor progression (TTP) and survival were calculated as days (d) from the date of CT scan. Sixty-nine tumors from 67 dogs were evaluated. Fifty-seven tumors were classified as carcinomas and 12 as non-carcinomas. Fifteen dogs had metastasis. All dogs were treated with surgical resection; 15 dogs received postoperative chemotherapy. Median tumor volume was 34cm3 (0.1 – 1196cm3). There was wide variation in first-order statistics. Mean Hounsfield units (HU) ratio (p = 0.022) and median mean HU ratio (p = 0.021) were significantly higher in carcinomas than non-carcinomas. Tumor sphericity was strongly correlated to volume (rs = 1.0) and mean HU was strongly correlated to median HU (rs = 0.9); other parameters were not correlated with each other. Median TTP and overall median survival time (MST) were 229 d and 322 d, respectively. MST was significantly longer (p = 0.0092) for carcinomas (357d) compared to non-carcinomas (56d). When carcinomas were considered separately, volume was significantly associated with TTP (p < 0.0001) and MST (p < 0.0001). Metastasis at diagnosis significantly decreased MST (78d versus 407d; p = 0.008). Further study of radiomic features in canine lung tumors is warranted, particularly given that it non-invasively provides additional tumor data.