A comparison of two methods for segmentation of functional volumes in radiotherapy planning of lung cancer patients. (3rd April 2021)

• Record Type: Journal Article
• Authors: Nyeng, T. B.; Møller, D. S.; Farr, K.; Kramer, S.; Khalil, A. A.; Grau, C.; Hoffmann, L.
• Is Part Of: Acta oncologica. Volume 60:Number 3(2021)
• Page Start: 353
• Page End: 360
• Publication Date: 2021-04-03
• Subjects: Functional imaging -- radiotherapy -- lung cancer -- pneumonitis; Oncology -- Periodicals; Cancer -- Treatment -- Periodicals; 616.992
• DOI: 10.1080/0284186X.2021.1877811
• Languages: English
• ISSNs: 0284-186X

Abstract

Abstract: Background: In radiotherapy (RT) of lung cancer, dose to functional lung (FL) volumes segmented with two different methods (perfusion SPECT (Q-SPECT) and 4D-CT (4D) ventilation (V)) have been shown to correlate with the incidence of radiation pneumonitis (RP). This study aims to compare the FL volumes identified by both methods. Material and methods: Thirty lung cancer patients had a 4D and Q-SPECT prior to treatment. Seventeen of these patients also had a ventilation SPECT (V-SPECT). FL sub-volumes were segmented automatically, using cut-off values. The volumes were compared in terms of overlap fraction (OF) relative to the minimal volume, and intersection fraction (IF) of the FL volume relative to the total lung volume ( V Lung ). Results: Cut-off values suggested in literature for Q-SPECT and 4D-V resulted in volumes differing in size by a median 18% [6%;31%], and a median OF and IF of 0.48 [0.23;0.70] and 0.09 [0.02;0.25], respectively. Segmenting volumes of comparable size of about 1/3 of V Lung (FL-m(1/3), m = method) resulted in a median OF and IF of 0.43 [0.23;0.58] and 0.12 [0.06;0.19], respectively. Twenty-five patients (83%) had a reasonable overlap between FL-Q(1/3) and FL-4D-V(1/3) volumes, with OF values above 0.33. IF increased significantly ( p = .036) compared to using fixed cut-off values. Similarly, volumes of comparable size of about 1/3 V Lung were produced for V-SPECT, and FL-Q(1/3), FL-V(1/3), and FL-4D-V(1/3) were compared. The overlaps and intersections of FL-V(1/3) with FL-Q(1/3) volumes were significantly ( p <.001) larger than the corresponding overlaps and intersections of FL-Q(1/3) with FL-4D(1/3) and FL-V(1/3) with FL-4D(1/3). Conclusion: The Q-SPECT and 4D-V methods do not segment entirely the same FL volumes. A reasonable overlap of the volumes along with the findings of other studies that both correlate to RP incidence, suggests that a combination of both volumes, e.g. using the IF, may be useful in RT treatment planning.