Comparative benefits of Ki and SUV images in lesion detection during PET/CT imaging

Danjie Cai; Yibo He; Haojun Yu; Yiqiu Zhang; Hongcheng Shi

doi:10.1186/s13550-024-01162-x

Comparative benefits of Ki and SUV images in lesion detection during PET/CT imaging

EJNMMI Res. 2024 Oct 16;14(1):98. doi: 10.1186/s13550-024-01162-x.

Authors

Danjie Cai^#^{1

2

3}, Yibo He^#^{1

2

3}, Haojun Yu^{1

2

3}, Yiqiu Zhang^{1

2

3}, Hongcheng Shi^{4

5

6}

Affiliations

¹ Department of Nuclear Medicine, Zhongshan Hospital, Fudan University, Shanghai, 200032, China.
² Nuclear Medicine Institute of Fudan University, Shanghai, 200032, China.
³ Shanghai Institute of Medical Imaging, Shanghai, 200032, China.
⁴ Department of Nuclear Medicine, Zhongshan Hospital, Fudan University, Shanghai, 200032, China. shi.hongcheng@zs-hospital.sh.cn.
⁵ Nuclear Medicine Institute of Fudan University, Shanghai, 200032, China. shi.hongcheng@zs-hospital.sh.cn.
⁶ Shanghai Institute of Medical Imaging, Shanghai, 200032, China. shi.hongcheng@zs-hospital.sh.cn.

^# Contributed equally.

PMID: 39412599
DOI: 10.1186/s13550-024-01162-x

Abstract

Background: Clinical application of the tracer net influx rate (Ki) imaging in PET/CT remains limited, due to a lack of evidence demonstrating the superiority of Ki images in lesion detection, and guidelines on when to utilize Ki images. This study aims to compare the benefits of Ki and standardized uptake value (SUV) images in lesion detection during PET/CT imaging. By analyzing the performance of both techniques in identifying tumor lesions, the study seeks to provide guidance for the clinical application of Ki images.

Results: This retrospective study included 134 patients with 244 pathologically confirmed lesions (200 malignant and 44 benign). Patients with a histopathological diagnosis received a weight-based ¹⁸F-FDG injection and underwent 60-min total-body PET/CT dynamic imaging. SUV images were reconstructed using data collected from the last 10 min of the scans. Ki images were generated using the Patlak methods with data from minutes 12-60. The background SUV_max, SUV_mean, SUV_SD, Ki_max, Ki_mean, and Ki_SD values were recorded. The signal-to-noise ratios of the SUV (SUV_SNR) and Ki (Ki_SNR) images were calculated. The lesion detection rate and sensitivity of the SUV and Ki images were evaluated. The lesion-detection rates were 97.7% (214/219) and 99.5% (218/219) for the SUV and Ki images, respectively (p = .22). Five false-negative lesions on the SUV images were true-positive on the Ki images (3 hepatic malignancies and 2 metastatic lymph nodes). The sensitivity (94.0% vs. 96.0%, p = .22), specificity (41.9% vs. 41.9%, p > .99), accuracy (84.4% vs. 86.1%, p = .61), positive predictive value (87.9% vs. 88.1%, p = .94), negative predictive value (60.0% vs. 69.2%, p = .47), and the area under the curve [0.68 (95% confidence interval, 0.61-0.73) vs. 0.69 (95% confidence interval, 0.62-0.74)] were similar in the SUV and Ki images (all p ≥ .10).

Conclusion: Ki images exhibit benefits in lesion detection compared to SUV images, particularly in organs with high background such as liver. The enhanced contrast provided by Ki imaging is recommended to clinically improve detection rates in such cases.

Keywords: Dynamic imaging; Ki image; Lesion detection rate; PET/CT.

Abstract

Grants and funding