Purpose: To propose a novel method to determine α/β from post radiation CT scans for different fractionations that does not rely on binary endpoint analysis. Methods: Pre‐and post‐RT (<6mo, 6‐12mo and >12mo) CT images from patients with lung cancer treated at 3 institutions, using conventional RT (Duke/North Carolina, VU Medical center Netherlands) and hypofractionated RT (VU and University of Colorado) were compared. Changes in normal lung tissue density were assessed and related to the regional RT dose (considered as a continuous variable) to generate a dose‐density response curves (DRC). Principal Component Analysis was used to characterize the deviation of each of the 449 DRCs to the mean DRC. In the low dose region (<50Gy in 2Gy), we hypothesized that the distributions for conventional (Conv‐fx) and hypofractionated (Hypo‐fx) DRCs corrected for biologically equivalent dose were similar. Since α/β adjusts the equivalent dose its optimal value maximized the similarity between the Conv‐fx‐based and Hypo‐fx‐based DRC distributions. Results: The 3 institution time‐aggregated data set returned an α/β of 3.7 (233 Conv‐fx DRCs / 216 Hypo‐fx). The data at < 6 mo was α/β of 2.7 (60 Conv‐fx/ 85 Hypo‐fx), and at > 6 mo was 3.9 at (173 Conv‐fx / 131 Hypo‐fx). Restricted to the two VU fractionation sets, α/β was 2.3 (12 Conv‐fx/ 50 Hypo‐fx) at < 6mo. Adding the CU data to the VU data, α/β ratio became 2.8 (12 Conv‐fx/102 Hypo‐fx) indicating relative Result stability with regard to sample size. To be consistent with other institutions, the Duke/North Carolina data needed to be corrected by the density value in non‐irradiated regions. Conclusion: The estimated α/β value of 3.7 based on CT density changes is endpoint independent and compatible with published values for clinical endpoints. Despite inter‐institution processing variations, values from different subsets were consistent suggesting tolerance to noise.