Generative adversarial network-based attenuation correction for Tc-99m-TRODAT-1 brain SPECT

Background: Attenuation correction (AC) is an important correction method to improve the quantification accuracy of dopamine transporter (DAT) single photon emission computed tomography (SPECT). Chang's method was developed for AC (Chang-AC) when CT-based AC was not available, assuming uniform attenuation coeffcients inside the body contour. This study aims to evaluate Chang-AC and different deep learning (DL)-based AC approaches on Tc-99m-TRODAT-1 brain SPECT using clinical patient data on two different scanners. Methods: Two hundred and sixty patients who underwent 99mTc-TRODAT-1 SPECT/CT scans from two different scanners (scanner A and scanner B) were retrospectively recruited. The ordered-subset expectation-maximization (OS-EM) method reconstructed 120 projections with dual-energy scatter correction, with or without CT-AC. We implemented a 3D conditional generative adversarial network (cGAN) for the indirect deep learning-based attenuation correction (DL-AC(mu)) and direct deep learning-based attenuation correction (DL-AC) methods, estimating attenuation maps (mu-maps) and attenuation-corrected SPECT images from non-attenuation-corrected (NAC) SPECT, respectively. We further applied cross-scanner training (cross-scanner indirect deep learning-based attenuation correction [cull-AC(mu)] and cross-scanner direct deep learning-based attenuation correction [call-AC]) and merged the datasets from two scanners for ensemble training (ensemble indirect deep learning-based attenuation correction [eDL-AC(mu)] and ensemble direct deep learning-based attenuation correction [eDL-AC]). The estimated mu-maps from (c/e)DL-AC(mu) were then used in reconstruction for AC purposes. Chang's method was also implemented for comparison. Normalized mean square error (NMSE), structural similarity index (SSIM), specific uptake ratio (SUR), and asymmetry index (%ASI) of the striatum were calculated for different AC methods. Results: The NMSE for Chang's method, DL-AC mu, DL-AC, cDL-AC mu, cDL-AC, eDL-AC mu, and eDL-AC is 0.0406 +/- 0.0445, 0.0059 +/- 0.0035, 0.0099 +/- 0.0066, 0.0253 +/- 0.0102, 0.0369 +/- 0.0124, 0.0098 +/- 0.0035, and 0.0162 +/- 0.0118 for scanner A and 0.0579 +/- 0.0146, 0.0055 +/- 0.0034, 0.0063 +/- 0.0028, 0.0235 +/- 0.0085, 0.0349 +/- 0.0086, 0.0115 +/- 0.0062, and 0.0117 +/- 0.0038 for scanner B, respectively. The SUR and %ASI results for DL-AC mu are closer to CT-AC, Followed by DL-AC, eDL-AC(mu), cDL-AC(mu), cDL-AC, eDL-AC, Chang's method, and NAC. Conclusion: All DL-based AC methods are superior to Chang-AC. DL-AC(mu) is superior to DL-AC. Scanner-specific training is superior to cross-scanner and ensemble training. DL-based AC methods are feasible and robust for Tc-99m-TRODAT-1 brain SPECT.

Automated summary

DL-AC(mu) method outperforms DL-AC and Chang-AC methods in Tc-99m-TRODAT-1 brain SPECT, with better correction accuracy. Scanner-specific training is superior to cross-scanner and ensemble training. DL-based AC methods are feasible and robust for Tc-99m-TRODAT-1 brain SPECT.