*3.6. Additional Value of Dual-Time-Point [18F]FDG Kinetic*

The implementation of dual-time-point parameters significantly improved the model fit. The most predictive parameters were <sup>Δ</sup>SUVpeak (log-likelihood: −42.66; <sup>χ</sup><sup>2</sup> difference = 7.11; *<sup>p</sup>* < 0.01) and RI-SUVpeak (log-likelihood: −43.44; <sup>χ</sup><sup>2</sup> difference = 5.20; *<sup>p</sup>* = 0.02; log-likelihood of the comparison model without these dual-time-point parameters: −46.21, *n* = 144).

Implementing the dual-time-point [18F]FDG kinetic parameter ΔSUVpeak in the MS lowered error rates in G2 tumors by one-third from 65.5% to 44.5% compared with the best single parameter SUVavg (Supplementary Figure S1).

Inclusion of ΔSUVpeak and RI-SUVpeak resulted in a slightly but not significantly increased discriminatory power (MS + ΔSUVpeak: AUC: 0.837; sensitivity: 79.3%; specificity: 75.7%) compared with the standard MS model (AUC: 0.820; sensitivity: 72.2%; specificity: 80.7%).

#### *3.7. Visual vs. Multiparametric Evaluation*

Specificity was set by the visual evaluation, and corresponding sensitivity was compared between visual and multiparametric LN evaluation using MS. Applying MS increased the overall sensitivity from 31.0% to 37.9% compared with the expert consensus at a set specificity of 98.3% (*n* = 144, prevalence: 20.1%), although the defined specificity was far from the Youden optimum of the MS (sensitivity: 79.3%; specificity: 75.6% at cut-off of 0.0042).

For G3 tumors, MS revealed a higher sensitivity (47.1% vs. 58.8%) compared with the human reader at a set specificity of 96.3% (*n* = 71, prevalence 23.9%), which was close to the Youden optimum (sensitivity of 76.4% at a specificity of 85.1%; cut-off: 0.0908).

For G2 LNM, using MS, resulted in an identical sensitivity of 8.3% at a set specificity of 100% (*n* = 73, prevalence: 16.4%). However, sensitivity increased from 8.3% to 83.3% if adjusted to the Youden optimum at a specificity of 72.1% (cut-off: 0.0435).

## **4. Discussion**

To our knowledge, this is the first prospective study analyzing the additional diagnostic value of a multiparametric [18F]FDG PET/MRI analysis compared with expert consensus reading for N-staging with histology as the gold standard in FIGO I/II cervical carcinomas. A multiparametric malignancy score was introduced, which integrates dualtime-point [18F]FDG kinetics and biopsy-based grading of the primary tumor in addition to established PET and MRI parameters. Using [99mTc]Tc-Nanocolloide for SLN labeling provided accurate transfer of LN positions via SLN SPECT/CT to PET/MRI, resulting in high data quality, which is a strength of this study.

Our results indicate that multiparametric analysis using the MS may double the sensitivity in LNM detection in FIGO I/II cervical cancer in G2 tumors compared with visual evaluation. As PET/MRI has already been shown to improve T- and M-staging, enhancing the accuracy in N-staging is the next big step in optimizing noninvasive staging for cervical carcinoma. This is of high clinical relevance, as surgical LN staging is currently the first step of surgery in advanced cervical cancer (in contrast to early cancer, where radical hysterectomy is usually the first step, followed by (sentinel-) LN dissection) [2]. Furthermore, preoperative assessment and evaluation of nodal involvement have a direct therapeutic impact as the presence of LNM leads to a change from radical hysterectomy to radiochemotherapy according to current guidelines [2].
