Validation of low-dose lung cancer PET-CT protocol and PET image improvement using machine learning. (January 2021)

• Record Type: Journal Article
• Authors: Nai, Ying-Hwey; Schaefferkoetter, Josh; Fakhry-Darian, Daniel; O'Doherty, Sophie; Totman, John J.; Conti, Maurizio; Townsend, David W.; Sinha, Arvind K.; Tan, Teng-Hwee; Tham, Ivan; Alexander, Daniel C.; Reilhac, Anthonin
• Is Part Of: Physica medica. Volume 81(2021)
• Page Start: 285
• Page End: 294
• Publication Date: 2021-01
• Subjects: Positron emission tomography (PET) -- Lung cancer -- Machine learning -- Lesion detection; SD Standard Dose -- LD Low-Dose -- ALD Actual Low-Dose -- SLD Simulated Low-Dose -- IQT Image Quality Transfer -- PET Positron Emission Tomography -- CT Computed Tomography -- VOI Volume of Interest -- GL Global Linear regression -- ST non-linear Single Tree regression -- RF non-linear Random Forest regression -- SUV Standardized Uptake Values -- NMSE Normalized Mean-Square-Error -- NMedSE Normalized Median-Square-Error -- STD Standard Deviation -- CNR Contrast-to-Noise Ratio -- AUROC Area Under Receiver Operating Characteristics Curve (Calculated parameters are italicized for ease of differentiation between names of images and metrics; Medical physics -- Periodicals; Biophysics -- Periodicals; Biophysics -- Periodicals; Imagerie médicale -- Périodiques; Radiothérapie -- Périodiques; Rayons X -- Sécurité -- Mesures -- Périodiques; Physique -- Périodiques; Médecine -- Périodiques; 610.153
• DOI: 10.1016/j.ejmp.2020.11.027
• Languages: English
• ISSNs: 1120-1797

Abstract

Highlights: Lesion detectability was not lost using proposed low-dose PET-CT protocol. Low-dose PET-CT protocol reduces radiation exposure to 43% of standard protocol. Machine-learning increased lesion contrast and reduced false positives. Machine-learning enables radiation exposure reduction to 34% of standard protocol. Abstract: Purpose: To conduct a simplified lesion-detection task of a low-dose (LD) PET-CT protocol for frequent lung screening using 30% of the effective PETCT dose and to investigate the feasibility of increasing clinical value of low-statistics scans using machine learning. Methods: We acquired 33 SD PET images, of which 13 had actual LD (ALD) PET, and simulated LD (SLD) PET images at seven different count levels from the SD PET scans. We employed image quality transfer (IQT), a machine learning algorithm that performs patch-regression to map parameters from low-quality to high-quality images. At each count level, patches extracted from 23 pairs of SD/SLD PET images were used to train three IQT models – global linear, single tree, and random forest regressions with cubic patch sizes of 3 and 5 voxels. The models were then used to estimate SD images from LD images at each count level for 10 unseen subjects. Lesion-detection task was carried out on matched lesion-present and lesion-absent images. Results: LD PET-CT protocol yielded lesion detectability with sensitivity of 0.98 and specificity of 1. Random forest algorithm with cubic patch size of 5 allowed further 11.7% reduction in the effective PETCT dose without compromising lesion detectability, but underestimated SUV by 30%. Conclusion: LD PET-CT protocol was validated for lesion detection using ALD PET scans. Substantial image quality improvement or additional dose reduction while preserving clinical values can be achieved using machine learning methods though SUV quantification may be biased and adjustment of our research protocol is required for clinical use.