The influence of digital PET/CT on diagnostic certainty and interrater reliability in [68Ga]Ga-PSMA-11 PET/CT for recurrent prostate cancer

Abstract

Objective: To investigate the impact of digital PET/CT on diagnostic certainty, patient-based sensitivity and interrater reliability.

Methods: Four physicians retrospectively evaluated two matched cohorts of patients undergoing [⁶⁸Ga]Ga-PSMA-11 PET/CT on a digital (dPET/CT n = 65) or an analogue scanner (aPET/CT n = 65) for recurrent prostate cancer between 11/2018 and 03/2019. The number of equivocal and pathological lesions as well as the frequency of discrepant findings and the interrater reliability for the two scanners were compared.

Results: dPET/CT detected more lesions than aPET/CT (p < 0.001). A higher number of pathological scans were observed for dPET/CT (83% vs. 57%, p < 0.001). The true-positive rate at follow-up was 100% for dPET/CT compared to 84% for aPET/CT (p < 0.001). The proportion of lesions rated as non-pathological as a total of all PSMA-avid lesions detected for dPET/CT was comparable to aPET/CT (61.8% vs. 57.0%, p = 0.99). Neither a higher rate of diagnostically uncertain lesions (11.5% dPET/CT vs. 13.7% aPET/CT, p = 0.95) nor discrepant scans (where one or more readers differed in opinion as to whether the scan is pathological) were observed (18% dPET/CT vs. 17% aPET/CT, p = 0.76). Interrater reliability for pathological lesions was excellent for both scanner types (Cronbach's α = 0.923 dPET/CT; α = 0.948 aPET/CT) and interrater agreement was substantial for dPET/CT (Krippendorf's α = 0.701) and almost perfect in aPET/CT (α = 0.802).

Conclusions: A higher detection rate for pathological lesions for dPET/CT compared with aPET/CT in multiple readers was observed. This improved sensitivity was coupled with an improved true-positive rate and was not associated with increased diagnostic uncertainty, rate of non-specific lesions, or reduced interrater reliability.

Key points: • New generation digital scanners detect more cancer lesions in men with prostate cancer. • When using digital scanners, the doctors are able to diagnose prostate cancer lesions with better certainty • When using digital scanners, the doctors do not disagree with each other more than with other scanner types.

Keywords: Molecular imaging; Nuclear medicine; Positron emission tomography; Prostate cancer.

Fig. 1

Total number of lesions identified by each reader across both scanners where (A) = aPET/CT and (D) = dPET/CT. Greater numbers of benign, equivocal, and pathological lesions were identified by all four readers combined at dPET/CT (p < 0.05)

Fig. 2

The rate of non-specific findings (NSF = detected lesions classified as non-pathological or equivocal, i.e. PSMA-RADS 1-3 as a proportion of all lesions detected)

Fig. 3

Rates of diagnostically uncertain lesions (equivocal) as a percentage of the total PSMA-avid lesions detected

Fig. 4

Patient-based sensitivity for all four readers (= scans rated as pathological, i.e. at least one PC lesion in PSMA-RADS category 4-5). Statistically significant results are marked with an asterisk (*)

Fig. 5

STARD flowchart showing the number of pathological scans where follow-up was available which were ultimately reported at follow-up to be true positives, and the non-pathological scans which were reported at follow-up to be false negatives

Fig. 6

Example images showing a benign ganglion with intensive PSMA-uptake (1, leftmost), equivocal uptake in a mediastinal lymph node (2, middle), and clearly pathological uptake (3, rightmost) in a pelvic lymph node. Top row fusion of PET and CT, bottom row. The appropriate window levels affording the best appreciation of the lesion are shown in the legend for each panel