To evaluate the contributory value of positron emission tomography (PET)-intravoxel incoherent motion (IVIM) magnetic resonance imaging (MRI) in the prediction of lymphovascular space invasion (LVSI) in patients with cervical cancer without lymphatic metastasis. A total of 90 patients with cervical cancer without signs of lymph node metastasis on PET/MRI were enrolled in this study. The tumours were classified into LVSI-positive (n = 25) and LVSI-negative (n = 65) groups according to postoperative pathology. The PET-derived parameters (SUVmax, SUVmean, metabolic tumour volume (MTV) and total lesion glycolysis (TLG)) and IVIM-derived parameters (ADCmean, ADCmin, Dmean, Dmin, f, D* and gross tumour volume (GTV)) between the two groups were evaluated using a Student's t test (Mann-Whitney U test for variables with a nonnormal distribution) and receiver operating characteristic (ROC) curves. The optimal combination of PET/MR parameters for predicting LVSI was investigated using univariate and multivariate logistic regression models and evaluated by ROC curves. The optimal cutoff threshold values corresponded to the maximal values of the Youden index. A control model was established using 1000 bootstrapped samples, for which the performance was validated using calibration curves and ROC curves. PET-derived parameters (SUVmax, SUVmean, MTV, TLG) and IVIMMRI-derived parameters (Dmin, ADCmin, GTV) were significantly different between patients with and without LVSI (P < 0.05). Logistic analyses showed that a combination of TLG and Dmin had the strongest predictive value for LVSI diagnosis (area under the curve (AUC), 0.861; sensitivity, 80.00; specificity, 86.15; P < 0.001). The optimal cutoff threshold values for Dmin and TLG were 0.58 × 10-3 mm2/s and 66.68 g/cm3, respectively. The verification model showed the combination of TLG and Dmin had the strongest predictive value, and its ROC curve and calibration curve showed good accuracy (AUC, 0.878) and consistency. The combination of TLG and Dmin may be the best indicator for predicting LVSI in cervical cancer without lymphatic metastasis.
Read full abstract