Key points are not available for this paper at this time.
Background Cervical cancer (CC) remains the second leading cause of cancer-related death in women, and the ability to accurately anticipate the presence or absence of lymphovascular space invasion (LVSI) is critical to maintaining optimal patient outcomes. The objective of this study was to establish and verify an MRI radiomics-based model to predict the status of LVSI in patients with operable CC. Methods The current study performed a retrospective analysis, with 86 patients in the training cohort and 38 patients in the testing group, specifically focusing on patients with CC. The radiomics feature extraction process included ADC, T2WI-SPAIR, and T2WI sequences. The training group data were used for the initial radionics-based model building, and the model predictive performance was subsequently validated using data from patients recruited in the experimental group. Results The development of the radiomics scoring model has been completed with 17 selected features. The study found several risk factors associated with LVSI. These risk factors included moderate tumor differentiation (P = 0.005), poor tumor differentiation (P = 0.001), and elevated combined sequence-based radiomics scores (P = 0.001). Radiomics scores based on predictive model, combined sequences, ADC, T2WI-SPAIR, and T2WI exhibited AUCs of 0.897, 0.839, 0.815, 0.698, and 0.739 in the training cohort, respectively, with corresponding testing cohort values of 0.833, 0.833, 0.683, 0.692, and 0.725. Excellent consistency was shown by the calibration curve analysis, which showed a higher degree of agreement between the actual and anticipated LVSI status. Moreover, the decision curve analysis outcomes demonstrated the medical application of this prediction model. Conclusion This investigation indicated that the MRI radiomics model was successfully developed and validated to predict operable CC patient LVSI status, attaining high overall diagnostic accuracy. However, further external validation and more deeper analysis on a larger sample size are still needed.
Ma et al. (Fri,) studied this question.