Reconstruction-informed and multidomain deep learning for generalizable CT-free attenuation correction in SPECT myocardial perfusion imaging.
Deep-learning attenuation correction shows near-perfect agreement with CT-based correction in SPECT imaging.
Reconstruction-informed and multidomain deep learning for generalizable CT-free attenuation correction in SPECT myocardial perfusion imaging.
Deep learning (deep learning) has shown promise in enabling attenuation correction (attenuation correction) for SPECT myocardial perfusion imaging (myocardial perfusion imaging) without relying on anatomical information or CT-derived attenuation maps (attenuation maps).
A dataset of 1058 SPECT/CT myocardial perfusion imaging scans using 99m Tc-Sestamibi from two centers was used for training (934 cases) and an external test set (124 cases).
Our proposed approach significantly outperformed direct and indirect methods in ablation comparison.
This model yielded mean relative absolute error percentage (MRAE%) values of 25.02 ± 23 (internal) and 26.31 ± 14 (external) for attenuation maps, and 11.72 ± 6.3 (internal) and 19.31 ± 4.9 (external) for attenuation correction SPECT images.
Organ-wise analysis showed region-wise MRAE% of 9.29 ± 6.5 (internal) and 17.28 ± 17 (external) in the attenuation maps domain, and 4.51 ± 4.3 (internal) and 9.53 ± 6.6 (external) in the attenuation correction domain.
Polar map analysis across 17 segments showed MRAE% of 5.04 ± 4.6 (internal) and 10.88 ± 7.3 (external).
This study demonstrated that our proposed reconstruction-informed and multidomain method utilizing multiple OSEM reconstruction inputs and jointly optimizing attenuation maps and attenuation correction losses substantially improved model performance in the indirect strategy.
The indirect method consistently outperformed the direct approach, and our model generalized well on external data, showing strong agreement with SPECT CTAC images in both quantitative and qualitative assessments.