Abstract
1538
Introduction: The possibility of reduced ionization dose of ultra-high-sensitivity total-body PET makes attenuation computed tomography (CT) a critical radiation burden in clinical applications. Artificial intelligence has shown the potential to generate PET images from non-attenuation corrected PET images. Our aim in this work is to develop a CT-free attenuation correction (AC) for a long field of view (FOV) PET scanner.
Methods: Whole body PET images of 165 patients scanned with two Biograph Vision PET/CT scanners, located in Shanghai and Bern, were used for the development and testing of the deep learning methods. The developed algorithm was tested on data of 10 patients scanned with a long axial FOV scanner, the Biograph Vision Quadra, in Bern). A generative adversarial network (GAN) was developed featuring a residual dense block, which enables the model to fully exploit hierarchical features from all network layers. Transfer learning was applied to a Biograph Vision Quadra dataset to leverage the performance. The normalized root mean squared error (NRMSE) and peak signal-to-noise ratio (PSNR), were calculated to evaluate the results generated by deep learning.
Results: The preliminary results showed that, the developed deep learning method achieved an average NRMSE of 0.4±0.3% and PSNR of 51.4±6.4 for the test on Biograph Vision and an average NRMSE of 1.0±0.3% and PSNR of 40.3±3.1 for the validation on Biograph Vision Quadra. After transfer learning, the model was able to achieve an average NRMSE of 0.5±0.4% and PSNR of 47.9±9.4 on Biograph Vision Quadra.
Conclusions: The developed deep learning method shows the potential for CT-free AC for a long axial FOV PET scanner, and transfer learning may accelerate this implementation. The CT-free AC PET images are undergoing clinical assessment and being compared with images obtained with conventional CT-based AC, by nuclear medicine physicians.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.