Magnetic Resonance Imaging and Diffusion Weighted Imaging-Based Histogram in Predicting Mesenchymal Transition High-Grade Serous Ovarian Cancer
Song-Qi Cai,Zhen-Yu Song,Min-Rong Wu,Wen-Wen Sun,Feng Wei,Hai-Ming Li,Jin-Wei Qiang,Yong-Ai Li,Jian Zhu,Jian-Jun Zhou,Meng-Su Zeng,Jing-Jing Lu
DOI: https://doi.org/10.1016/j.acra.2022.06.021
IF: 5.482
2023-01-01
Academic Radiology
Abstract:Rationale and Objectives: To investigate the value of magnetic resonance imaging (MRI) including diffusion-weighted imaging (DWI) find-ings in predicting mesenchymal transition (MT) high-grade serous ovarian cancer (HGSOC).Materials and Methods: Patients with HGSOC were enrolled from May 2017 to December 2020, who underwent pelvic MRI including DWI (b = 0,1000 s/mm2) before surgery, and were assigned to the MT HGSOC or non-MT HGSOC group according to histopathology results. Clinical characteristics and MRI features including DWI-based histogram metrics were assessed and compared between the two groups. Univariate and multivariate analyses were performed to identify the significant variables associated with MT HGSOC -these vari-ables were then incorporated into a predictive nomogram, and ROC curve analysis was subsequently carried out to evaluate diagnostic performance.Results: A total of 81 consecutive patients were recruited for pelvic MRI before surgery, including 37 (45.7%) MT patients and 44 (54.3%) non-MT patients. At univariate analysis, the features significantly related to MT HGSOC were identified as absence of discrete primary ovarian mass, pouch of Douglas implants, ovarian mass size, tumor volume, mean, SD, median, and 95th percentile apparent diffusion coefficient (ADC) values (all p < 0.05). At multivariate analysis, the absence of discrete primary ovarian mass {odds ratio (OR): 46.477; p = 0.025}, mean ADC value < 1.105 (OR: 1.023; p = 0.009), and median ADC value < 1.038 (OR: 0.982; p = 0.034) were found to be inde-pendent risk factors associated with MT HGSOC. The combination of all independent criteria yielded the largest AUC of 0.82 with a sensi-tivity of 83.87% and specificity of 66.67%, superior to any of the single predictor alone (p < 0.012). The predictive C-index nomogram performance of the combination was 0.82.Conclusion: The combination of absence of discrete primary ovarian mass, lower mean ADC value, and median ADC value may be helpful for preoperatively predicting MT HGSOC.