CT perfusion heterogeneity as a prognostic biomarker in NSCLC

Purpose: To explore correlation between features extracted by CT perfusion (CTp) colorimetric maps and overall survival, to assess whether they could work as a prognostic image-based biomarker. Methods and Materials: 16 axial CTp examinations, referring to as many patients (age range 36-79 years) with NSCLC, were considered and survival data were included in the study. Median overall survival was calculated at 8 months. Blood flow (BF) values were computed according to the maximum slope method and values undergoing high fitting errors have being automatically removed. Seven different features are computed on BF maps and k-means clustering algorithm was used to assess their correlation with overall survival (OS). Results: Two texture features, uniformity (U) and entropy (E), this reflecting irregularity and inhomogeneity, used together result to be the most effective prognostic factors for OS. The group of 11 patients with E>7.0 and U<0.01 shows OS≤8 and are clearly linearly separated from the second group of 5 patients having OS>8. Conclusions: These results confirm the common expectation that higher heterogeneity correlates with higher aggressiveness and this heavily reflects on the OS parameter. In addition, while often predictive features are looked for original (HU) CT images, results prove a strong correlation with OS by simple features computed on perfusion maps, after that erroneous perfusion values have been automatically removed. This finding also implicitly considers the measure of hemodynamic heterogeneity of tumour as a relevant and objective prognostic factor for OS and represents a promising approach to clinical utilization of CTp.