*4.2. Dual-Time-Point [18F]FDG PET*

In contrast to previous studies, we introduced a short-period dual-time-point imaging protocol using an interval of 30 min instead of 1–2.5 h [25,27,28], which enables continuous scanning without the need for repositioning patients. Even for this short time interval, a significant increase in tumoral [18F]FDG uptake was found—calculated as RI as proposed by Nogami et al. [25]—which was a significant independent predictor of malignancy. Furthermore, it was shown for the first time that the increase in [18F]FDG uptake over time was only significant for G3 LNM but not for G2 LNM. G2 LNM presented with a slight decrease in SUV analogous to the decline of blood pool activity, which might be explained by lower metabolic activity and tumor cell density [29].

Blood pool correction was crucial for dual-time-point dynamic measurements owing to a decreasing blood pool and increasing scatter correction artefacts caused by rising activity concentrations in the bladder. Although the dual-time-point kinetics were a significant factor, delayed PET images did not outperform the early scans. This might be due to LNM with increasing SUV dynamics already showing increased uptake on early PET scans.

#### *4.3. Experts vs. Malignancy Score*

Visual evaluation of LNM by expert readers was highly specific but accompanied by poor sensitivity, which runs counter to the principle of presurgical screening.

By using the MS with a cut-off value at the Youden optimum, the sensitivity could be improved substantially, especially in G2. The moderate loss of specificity would be acceptable, as false-positive pelvic LN are re-evaluated during surgery.

The sensitivity of visual evaluation was even lower than described in previous studies (31% vs. 45–88%) [30,31]. This might be due to our cohort of solely early-stage carcinomas and the higher number of micrometastases detectable by ultrastaging, as discussed above.

#### *4.4. Limitations*

The results presented here only pertain to G2 and G3 tumors as no LNM occurred in the small number of G1 tumors in our cohort.

The smaller FOV of the delayed [18F]FDG PET/MRI scan limited the dual-time-point analysis to pelvic LN.

In order to avoid further strain on the information extracted from the data, a listwise exclusion of cases was applied throughout the analyses; however, this resulted in a varying number of LNM in the results.

Furthermore, the multiparametric evaluation was based on histology and, therefore, was performed retrospectively in contrast to prospective reading of the experts. Consequently, the multiparametric analysis was subjected to accommodation of random effects to keep the diagnostic performance of MS comparable to the expert reading and other cohorts.

Prior to a broader clinical application, the presented MS should be validated prospectively in a comparable setting, which is planned for the second half of this ongoing clinical trial.

#### **5. Conclusions**

G2 vs. G3 tumor grade was identified as a crucial factor for limited visual detectability of LNM on [18F]FDG PET/MRI in early cervical carcinoma.

Multiparametric evaluation of dual-time-point [18F]FDG PET/MRI has the potential to considerably improve the accuracy of LNM detection and to extend sufficient N-staging also to G2 tumors.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jcm11174943/s1, Figure S1. Error rates for the detection of lymph node metastases by single parameter and multiparametric malignancy score in dependence of tumor grading. The multiparametric malignancy score (MS) (D) lowers the error rate (ER = false positive rate FNR + false negative rate FPR) about 5 percentage points and stabilizes the ER over a wider range compared to the best single parameter SUVeavg (A). Implementing dual timepoint kinetics (E) further enhances this effect and lowers the summed error rate ER by another 2 percentage points compared to the standard MS (D). In G2 tumors, this effect is most evident with a significant reduction in summed error rate ER of 21 percentage points from 65.5% (C) to 44.5% (F). Grade of primary tumor has an huge impact on detectability of lymph note metastases with an doubling of error rates in G2 LNM (C) compared to G3 LNM (B) with a sharp increase in FNR starting at an SUV of 1. Table S1. MRI imaging parameters. Table S2. Survey table of AUC analysis of the dual-time-point PET/MRI

parameters for G2 and G3 cervical carcinoma. No lymph node present in G1 tumors. Table S3. Effect of the tumor grade on PET/MR parameters of lymph node metastases.

**Author Contributions:** Conceptualization, M.W.; data curation, M.W. and S.H.; formal analysis, M.W. and J.J.; funding acquisition, S.K., K.N. and C.L.F.; investigation, M.W. and H.D.; methodology, M.W., J.J., C.L.F. and H.D.; project administration, M.W., S.K. and C.L.F.; resources, S.K., K.N., S.Y.B. and C.L.F.; software, J.J.; supervision, K.N., S.Y.B. and H.D.; visualization, M.W.; writing – original draft, M.W.; writing—review & editing, S.K., J.J., S.U., F.S., S.H., K.N., S.Y.B., C.L.F. and H.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy (EXC 2180–390900677).

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of University Hospital Tübingen (protocol code 173/2015BO01) on 28 October 2015.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to data protection regulations.

**Acknowledgments:** Special thanks to technicians Carsten Groeper, Hans Volz, and Gerd Zeger for the enthusiastic and accurate performance of PET/MRI examinations. We acknowledge the financial support for publication fees by Open Access Publishing Fund of the University of Tübingen.

**Conflicts of Interest:** The authors declare no conflict of interest.
