Evaluation of various apparent diffusion coefficient measurement techniques in pre-operative staging of early cervical carcinoma

Abstract

PurposeDW-MRI parameters such as ADC hold the potential for more reliable staging of cervical cancer. We compared 2D region of interest (ROI) measurement techniques to 3D tumor analysis in the evaluation of ADC for cervical cancer. Secondly, we evaluated the utility of ADC for assessing parametrial and/or lymph node involvement. MethodThis prospective patient cohort registered cervical cancer patients who underwent pre-operative MRI with T1, T2W, and axial DWI. Retrospectively, two observers independently and blindly scored mean, minimum, and maximum ADC using three methods: a) 3D-Tumor analysis, b) single freehand ROI (2D-Slice), and c) single circular ROI (2D-Circle). Another observer scored parametrial and lymph node involvement on T1/T2W sequences. Parametrial and/or lymph node involvement were determined by surgical-pathologic results. The diagnostic performance of ADC for predicting the latter was evaluated by ROC curve, uni and multivariate analyses. ResultsOf 58 included patients, parametrial and/or lymph node involvement was found in 9 and 11 patients, respectively. Mean ADC (ADCMean) was least dependent on ROI technique, with interobserver variability (ICC: 0.88–0.90) and linear correlation (Pearson’s r: 0.95–0.96). To the contrary, minimal and maximal ADC were significantly influenced by 2D-ROI techniques. ADCMean was shown to be an independent predictor of parametrial (AUC: 0.80–0.86; OR: 16, 1.4–178) and/or lymph node involvement (AUC: 0.74–0.79; OR: 5.1, 1.1–24). ConclusionsSingle ROI measurements are a reliable method for determining ADCMean in cervical cancer. Second, ADCMean serves as a potential parameter for prediction of parametrial involvement prior to radical hysterectomy.