Composing a dictionary of lesions showing SUVmax greater than 20: an exhaustive, computer-assisted study

Introduction: FDG-PET/CT is an essential diagnostic tool for lesion characterization, staging, treatment monitoring, and treatment response evaluation in clinical oncology. When interpreting images and making diagnosis for FDG-PET/CT, the maximum standardized uptake value (SUVmax) is a useful semiquantitative index for differential diagnosis in daily practice, although not all the lesions can be discriminated clearly by SUVmax cut-off. In our practice, we rarely encounter such lesions that have SUVmax greater than 20, while many lesions, including malignant tumors, benign tumors, and inflammatory lesions, show SUVmax smaller than 10. Some literatures have reported that aggressive lymphoma (e.g., diffuse large B cell lymphoma) often showed extremely high SUVmax. However, to our knowledge, there are no thorough reports that investigated lesions with SUVmax ≥ 20. Recently, artificial intelligence (AI) has made remarkable progress, and we thought that such automated techniques could be used to “Compose a Dictionary” for high SUVmax lesions. Thus, in this single-center, retrospective, computer-assisted study, we exhaustively extracted lesions of SUVmax≥20 from 12-month FDG-PET/CT images, and summarized the lesion locations and diagnoses.

Methods: We investigated a total of 2386 whole-body FDG-PET/CT images whose axial field-of-view longer than 80 cm acquired with 3 different scanners at our institute in 2019. For each image, we applied the following automated and human processes (Fig 1). As the Step 1, the attenuation-corrected whole-image was segmented using a cut-off of SUVmax≥20 automatically. SUVmax of each segment was measured and displayed in maximum intensity projection (MIP) of the anterior-posterior (AP) and the right-left (RL) directions. In the Step 2, a nuclear medicine physician visually evaluated all the MIP images and excluded the images when the segmentation was found only in the brain or the urinary tracts. Then, in the Step 3, the physician further evaluated all the segments for lesion location and characteristics (malignant or non-malignant) by viewing cross sectional images of PET/CT. All the lesions that passed the 3 steps were fully evaluated by electric medical record (EMR).

Results: A total of 1843 images showed at least one SUVmax≥20 segments (Step 1), among which 237 images showed segments other than brain, bladder and kidney (Step 2). Non-malignant and malignant findings were found in 110 and 133 cases, respectively (i.e., 6 cases showed both non-malignant and malignant findings). Non-malignant findings included accumulation in the injection site, the ureter, or the urine catheter (n=36), physiological accumulation in the pharynx or the intestinal tract (n=22) or in the myocardium (n=43), and accumulation in benign lesions (n=6). Physiological accumulation was often symmetrical in the head-and-neck region and continuous in the intestinal tract, which allowed differentiation. Although the heart was a common site for physiological accumulation, 7 of 43 patients had heart disease. It was difficult to differentiate between pathological and physiological accumulation in the heart. Of malignant findings, malignant lymphoma (n=36) was most common, followed by head-and-neck cancer (n=34) and lung cancer (n=32). As previously reported, diffuse large B cell lymphoma was the most common (74%) among malignant lymphoma showing SUVmax≥20, but surprisingly, two cases of follicular lymphoma (i.e., indolent lymphoma) presented with SUVmax≥20. Adenocarcinoma was the most common type of lung cancer (n=12, 37.5%). Squamous cell carcinoma was the second most common (n=8, 25%). In this study, none of small cell carcinoma showed SUVmax≥20.

Conclusions: In this single-center, retrospective, computer-assisted study, we exhaustively investigated all the lesions of SUVmax≥20 on FDG-PET/CT and clarified the frequency of each pathological type. The dictionary we composed in this study will contribute to daily interpretation of FDG/PET-CT.

• Download figure
• Open in new tab
• Download powerpoint