CT-based 3D quantification of relative pulmonary lobe perfusion from SPECT in preoperative lung cancer patients: Comparing a manual and a semi-automatic approach

Objectives Quantification of relative pulmonary lobar perfusion using scintigraphy is an established procedure to estimate the expected lung function after pulmonary surgery e.g. in bronchial carcinoma. Using SPECT-CT and a 3D-technique, exact pulmonary lobe perfusion fraction can be calculated based on patient anatomy and not only estimated. In a previous study we showed that a 3D-technique based on a time-consuming manual segmentation of the pulmonary lobes using the CT with thereafter VOI transfer to the perfusion SPECT studies produced distinctly different results compared to the old fashioned planar perfusion approach using rectangular ROIs projected onto the anterior and posterior views. Recently a commercial software package “HERMES Hybrid 3D - Lung Lobe Quantification” was introduced, which allows a quick semi-automatic segmentation of lung CTs with automatic transfer of the resulting VOIs to the corresponding perfusion SPECT-CT study and direct calculation of the relative lobar perfusion fraction. The aim of our study was to compare the results of the time-consuming manual segmentation approach (man) with the newer semi-automatic algorithm (semi) in a preoperative setting in lung cancer patients with borderline lung function.

Methods 30 lung cancer patients referred to routine pre-operative relative lobar perfusion quantification were included. Perfusion SPECT-CT (low-dose-CT) was done p.i. of about 150 MBq Tc-99m-MAA (SymbiaT, Siemens). Ventilation SPECT was done before perfusion SPECT to exclude paraneoplastic PE. An iterative Flash3D algorithm was used to reconstruct the SPECT data. The manual 3D segmentation approach was based on the low-doseCT using PMOD, with the resulting VOIs transferred to the coregistered perfusion SPECT data in order to calculate the relative lobar perfusion fractions. The semi-automatic approach used “HERMES Hybrid 3D - Lung Lobe Quantification” to produce relative lobar perfusion fractions based on the matched low-doseCT and diagnostic CT (a diagnostic CT is available in virtually all patients with lung cancer). A correlation analysis of the manual vs. semi-automatic results was done

Results The semi-automatic approach took 5-10 minutes for a complete evaluation whereas the manual segmentation and calculation took about 45 minutes. Relative lobar perfusion values were in lung cancer patients quite similar for both approaches: right upper lobe 14.8±6.6% (man) vs 16.5±8.5% (semi), middle lobe 13.7±5.0% (man) vs 11.6±4.9% (semi), right lower lobe 24.6±7.6% (man) vs 24.8±9.9% (semi), left upper lobe 24.0±8.4% (man) vs 24.0±9.0% (semi), left lower lobe 22.9±5.7% (man) vs 22.9±6.4% (semi). When correlating all relative perfusion values the slope of the regression line (semi-automatic vs. manual) was 1.05 with a correlation coefficient of 0.92.

Conclusions Excellent correlation was found for the relative lobar perfusion values resulting from the semi-automatic approach using “HERMES Hybrid 3D - Lung Lobe Quantification” and the manual approach using PMOD. Therefore, the much more accurate 3D-technique can now overcome the old fashioned planar method without time-consuming manual analysis. Hence the 3D-approach is now suitable for routine use. In larger clinical trials, it has to be tested whether there are prognostic differences between both 3D-approaches, but taking the good concordance in mind we do not believe so.