A Multicenter Study on Preoperative Assessment of Lymphovascular Space Invasion in Early-Stage Cervical Cancer Based on Multimodal MR Radiomics

Abstract

Background: As lymphovascular space invasion (LVSI) was closely related to lymph node metastasis and prognosis, the preoperative assessment of LVSI in early-stage cervical cancer is crucial for patients.

Purpose: To develop and validate nomogram based on multimodal MR radiomics to assess LVSI status in cervical cancer patients.

Study type: Retrospective.

Population: The study included 168 cervical cancer patients, of whom 129 cases (age 51.36 ± 9.99 years) from institution 1 were included as the training cohort and 39 cases (age 52.59 ± 10.23 years) from institution 2 were included as the external test cohort.

Field strength/sequence: There were 1.5 T and 3.0 T MRI scans (T1-weighted imaging [T1WI], fat-saturated T2-weighted imaging [FS-T2WI], and contrast-enhanced [CE]).

Assessment: Six machine learning models were built and selected to construct the radiomics signature. The nomogram model was constructed by combining the radiomics signature with the clinical signature, which was then validated for discrimination, calibration, and clinical usefulness.

Statistical tests: The clinical characteristics were compared using t-tests, Mann-Whitney U tests, or chi-square tests. The Spearman and LASSO methods were used to select radiomics features. The receiver operating characteristic (ROC) analysis was performed, and the area under the curve (AUC), accuracy, sensitivity, and specificity were calculated.

Results: The logistic regression (LR) model performed best in each sequence. The AUC of CE-T1-T2WI-combined was the highest in the LR model, with an AUC of 0.775 (95% CI: 0.570-0.979) in external test cohort. The nomogram showed high predictive performance in the training (AUC: 0.883 [95% CI: 0.823-0.943]) and test cohort (AUC: 0.830 [95% CI: 0.657-1.000]) for predicting LVSI. Decision curve analysis demonstrated that the nomogram was clinically useful.

Data conclusion: Our findings suggest that the proposed nomogram model based on multimodal MRI of CE T1WI-T2WI-combined could be used to assess LVSI status in early cervical cancer.

Evidence level: 4.

Technical efficacy: Stage 2.

Keywords: cervical cancer; lymphovascular space invasion; machine learning; magnetic resonance imaging; radiomics.