Patients
In 2016 a total number of 96 patients was referred to our department for routine V/P-SPECT/CT to predict postoperative residual lung function prior to surgical treatment or endoscopic procedures. 74 patients suffered from pulmonary tumours and 22 from emphysema or COPD. Since the study focuses on lung tumours the data of the 22 patients with emphysema or COPD as indication for the procedure were not included into the analysis.
Thirteen patients were excluded from this evaluation study due to the following reasons: previous pulmonary surgery (7 patients), lack of a timely diagnostic CT (2 patients), present tracheal cannula (2 patients), diagnostic CT in prone positioning (1 patient) and impossibility of segmentation of the diagnostic CT by the software algorithm (1 patient).
The remaining group of 61 patients comprised of 39 men and 22 women with a mean age of 70.4 ± 8.1 years (range 52–87 years).
Diagnostic CT
Sixty-one full-dose CTs were available from different institutions (53 of those CTs were with contrast media). Since some patients brought their CT with them not in all cases tube voltage, CTDIvol or DLP were available. Therefore we report the mean values from the available data. Mean tube voltage (for 61/61 pts) was 119.0 ± 5.4 kV, mean CTDIvol (for 53/61 pts) was 8.22 ± 4.60, mean DLP (for 53/61 pts) was 341.1 ± 232.9 mGy*cm, resulting mean dose (53/61 pts) was 5.58 ± 3.29 mSv.
Lung perfusion SPECT/CT
Perfusion SPECT/CT was performed using a double-head gamma camera (Siemens SymbiaT, Siemens Medical Systems, Hofmann Estates, IL, USA) equipped with low-energy high-resolution collimators. Before perfusion scanning, a ventilation scan was acquired in each patient to exclude relevant pulmonary embolism. A minimum fourfold higher count rate was guaranteed for the perfusion compared to the ventilation scan.
SPECT acquisition started after intravenous injection of about 160 MBq of Tc-99m labelled macroaggregated albumin (MAA) in upright position. Upright position was chosen since all pulmonary function tests – like spirometry - are done in upright position and we intend to predict the postoperative outcome for the patients in daily life in upright position. A 360°-acquisition in supine position with 32 projections, 8 s per projection and a matrix of 128 × 128 was used. Finally, a low-dose CT was performed (5 mm slice thickness; 110 kV). For the low-dose CT mean current was 22.7 ± 5.9 mAs, mean CTDIvol was 1.32 ± 0.34, mean DLP was 40.5 ± 11.3 mGy*cm, resulting mean dose was 0.66 ± 0.16 mSv.
During SPECT/CT acquisition all patients were allowed to breathe shallowly; no respiratory gating was performed.
Images were reconstructed without any information from the low-dose CT using 3D iterative reconstruction with scatter correction (Flash 3D, 4 iterations and 8 subsets) and were then transferred to a Siemens syngo workstation with MI application software (Siemens Medical).
3D quantification of lobar perfusion
For this retrospective study two different approaches concerning the CT based quantification of lung perfusion were pursued, evaluation time needed was counted for all quantifications.
In the hybrid approach SPECT (see Fig. 1) perfusion data were quantified after low-dose CT from the SPECT/CT camera was used for coregistration of the externally performed diagnostic CT.
In the non-hybrid approach (see Fig. 1) only perfusion SPECT data from the SPECT/CT camera and an externally performed diagnostic full-dose CT were used for quantification, the low–dose CT was omitted in this approach as if no hybrid equipment was available.
As a first step in the quantification process perfusion SPECT, co-acquired low-dose CT data and diagnostic CT were transferred to commercially available software (HERMES Hybrid 3D – Lung Lobe Quantification, HERMES Medical Solutions, Stockholm, Sweden).
The software allowed the hybrid and non-hybrid approaches to be processed using the same step by step wizard workflow. The coregistration was performed automatically by the software. The hybrid approach coregistered the diagnostic CT with the low-dose CT whilst the non-hybrid approach coregistered the diagnostic CT with the SPECT perfusion. The software used a rigid (translation only) mattes mutual information algorithm for both approaches. The coregistration algorithm parameters were identical for both approaches. The manual uniform scaling adjustment was used to compensate for breathing mis-match of the diagnostic CT. For the hybrid method 37 cases out of 61 had no or only little manual adjustments, 24 cases out of 61 had relevant manual adjustments. For the non-hybrid method 29 cases out of 61 had no or only little manual adjustments, 32 cases out of 61 had relevant manual adjustments.
The right and left lung volumes were computed from the diagnostic CT using an automatic segmentation algorithm. The lung volumes were split into lobar regions by marking 5 points in 3 slices for each fissure. The resulting lobar regions were automatically transferred to the SPECT perfusion and the relative distribution of perfusion per lobar region was computed (workflow see Fig. 2).
Data analysis
Statistical analyses were done using Origin 6.1 G (OriginLab Corp., Northampton, England) and MedCalc (MedCalc Software, Ostend Belgium). Mean values ± standard deviation of relative lobar perfusion were calculated from the hybrid approach and from the non-hybrid approach. Additionally mean processing times were calculated for both approaches. Data were tested for significant differences using the t-test for paired samples. A p-value < 0.05 after Bonferroni correction for multiple comparisons was considered significant. Correlation and especially Bland Altman analyses (Bland & Altman, 1986) of the resulting relative lobar perfusion between both approaches (hybrid approach vs. non-hybrid approach) were performed to test for good agreement of the results. An estimation for the required sample size was done, yielding a minimum sample size of 111 pairs when setting type I error to 0.05 and type II error to 0.2 with an expected mean of the difference of 0, expected SD of differences of 2 and a maximum allowed difference between both methods of 5. Intraclass correlation coefficient was calculated.