Deep learning based denoising algorithm on ¹⁸F-FDG PET/CT for detecting hepatic metastasis

¹⁸F-fluorodeoxyglucose positron emission tomogra- phy/computed tomography (¹⁸F-FDG PET/CT) is a widely used imaging modality for the detection of liver metastases. Recently, deep neural networks (DNNs) have demonstrated excellent performance in image denoising. This study aimed to assess the performance of a deep learning-based denois- ing algorithm, the Advanced Intelligent Clear-IQ Engine, for ¹⁸F-FDG PET/CT images in the detection of liver metastases. A total of 14 histopathologi- cally proven cancer patients with liver metastases who underwent ¹⁸F-FDG PET/CT between May 2020 and August 2022 were included in the study. The number and size of liver metastases were recorded to compare the lesion detectability between deep learning-based (DL) images and conventional Gaussian-filtered (GF) images. Quantitative anal- ysis was performed using metrics including the contrast-to-noise ratio (CNR) of the lesion and the signal-to-noise ratio (SNR) of the liver. DL images detected 113 out of 168 lesions. The respective number for GF images was 99. Fourteen lesions that were missed in GF images were detected in DL images. All of these 14 lesions were with a diameter ≤ 10mm. In quantitative analysis, DL images showed significantly higher CNR and SNR than GF images. DL images improved lesion detectability, particu- larly for cases with small lesions.