Maintaining SUV accuracy in low-count PET with PETfectior: a deep learning denoising solution.
Authors
Affiliations (1)
Affiliations (1)
- Fundación Centro Diagnóstico Nuclear (FCDN), Buenos Aires C1417CVE, Argentina.
Abstract
The diagnostic image quality of positron emission tomography (PET) acquisitions strongly depends on the administered radiotracer activity and acquisition time. However, keeping these variables as low as possible is desired to reduce patient radiation exposure and the radiopharmaceutical related costs of PET examinations. PETfectior is an artificial intelligence-based software that processes PET scans and increases the signal-to-noise ratio, obtaining high-quality images from low-count-rate images. In this study, we perform an initial clinical validation of PETfectior on images acquired with half of the counting statistics required to meet the most recent European Association of Nuclear Medicine quantitative standards for<sup>18</sup>F-FDG PET, evaluating lesion detectability, quantitative accuracy and image quality. 258 patients referred for<sup>18</sup>F-FDG PET/computed tomography (CT) were prospectively included. The standard-of-care scans (100% scans) were acquired and reconstructed according to EARL standards 2. Besides, half-counting-statistics versions of these images were obtained from list-mode data and processed with PETfectior (50% + PETfectior scans). All the oncologic lesions were segmented on both versions of the PET/CT images either manually or automatically and lesion detectability was evaluated. The SUV<sub>max</sub>of the lesions was measured and the quantitative concordance of 50% + PETfectior and 100% images was evaluated. Subjective image quality was visually assessed by two experienced physicians. 1649 lesions were detected in a total of 198 studies. The 50%+ PETfectior images showed high sensitivity for lesion detection (99.9%) and only one false positive was detected. Quantitative concordance of SUV<sub>max</sub>measured on 100% and 50% + PETfectior images was within ±12.5% (95% limits of agreement), with a bias of -1.01%. No difference in image quality between 50% + PETfectior and 100% images was found. We conclude that PETfectior can safely be used in the clinical practice with<sup>18</sup>F-FDG at half counting statistics, with high sensitivity and specificity, low quantitative bias and high subjective image quality, at least for the scanners evaluated in this study.