The local coefficient of variation: a quantitative index to measure the global coherence of perfusion values in CT studies of lung cancer

Purpose: The coefficient of variation (CV) is often used to assess the reliability of CT perfusion maps. This measure is “global” and neglects the heterogeneity of lesions. This work presents a novel quantitative index overcoming this limitation. Methods and Materials: 12 patients (36-81 y.o.) with NSCLC underwent axial CT perfusion in 15 examinations. Blood flow maps were computed before and after motion correction according to the Maximum-Slope method. The “global” coefficient of variation (gCV) was computed for each map. Besides, each voxel was assigned the value of a local coefficient of variation (lCV) computed in a small window around it, so to keep and gather information regarding local heterogeneity. Results for each lesion are also shown in a colorimetric map and the average lCV is the index to be compared with gCV. t-test was performed to assess statistical significance (p<0.01). Results: Anatomical mismatch between scans jeopardises the measure of perfusion parameters this yielding, in general, worse values of gCV and lCV parameters. However, gCV fails in recognizing two maps coming from as many uncorrected examinations as being worse than those achieved after motion compensation. On the contrary, the averaged lCV values always result smaller for motion corrected maps. Conclusion: Several clinical considerations are based on CV being a valid index to assess the reliability of perfusion values, and it may fail. The average lCV could replace gCV in the studies needing to assess the quality of perfusion maps and, ultimately, their reliability also including apparent inconsistencies caused by tumour heterogeneity.