Automated detection of lung cancer at ultralow dose PET/CT by deep neural networks – Initial results. (December 2018)

• Record Type: Journal Article
• Authors: Schwyzer, Moritz; Ferraro, Daniela A.; Muehlematter, Urs J.; Curioni-Fontecedro, Alessandra; Huellner, Martin W.; von Schulthess, Gustav K.; Kaufmann, Philipp A.; Burger, Irene A.; Messerli, Michael
• Is Part Of: Lung cancer. Volume 126(2018)
• Page Start: 170
• Page End: 173
• Publication Date: 2018-12
• Subjects: Lung cancer -- Artificial intelligence -- Deep learning -- PET/CT -- Low dose; Lungs -- Cancer -- Periodicals; Lung Neoplasms -- Abstracts; Lung Neoplasms -- Periodicals; Poumons -- Cancer -- Périodiques; Lungs -- Cancer; Periodicals; Electronic journals; Electronic journals; 616.99424
• DOI: 10.1016/j.lungcan.2018.11.001
• Languages: English
• ISSNs: 0169-5002

Abstract

Highlights: Clinical standard PET images and images with thirtyfold reduced dose were assessed. Neural networks had a high sensitivity (91.5%–95.9%) for lung cancer detection. Machine learning may be helpful in the detection of lung cancer in low dose FDG-PET. Abstract: Objectives: We evaluated whether machine learning may be helpful for the detection of lung cancer in FDG-PET imaging in the setting of ultralow dose PET scans. Materials and methods: We studied the performance of an artificial neural network discriminating lung cancer patients ( n = 50) from controls ( n = 50) without pulmonary malignancies. A total of 3936 PET slices including images in which the lung tumor is visually present and image slices of patients with no lung cancer were exported. The diagnostic performance of the artificial neural network based on clinical standard dose PET images (PET100% ) as well as with a tenfold (PET10% ) and thirtyfold (PET3.3% ) reduced radiation dose (∼0.11 mSv) was assessed. Results: The area under the curve of the deep learning algorithm for lung cancer detection was 0.989, 0.983 and 0.970 for standard dose images (PET100% ), and reduced dose PET10%, and PET3.3% reconstruction, respectively. The artificial neural network achieved a sensitivity of 95.9% and 91.5% and a specificity of 98.1% and 94.2%, at standard dose and ultralow dose PET3.3%, respectively. Conclusion: Our results suggest that machine learning algorithms may aid fully automated lung cancer detection even at very low effective radiation doses of 0.11 mSv. Further improvement of this technology might improve the specificity of lung cancer screening efforts and could lead to new applications of FDG-PET.