Abstract: Abstract: Objective　To evaluate the values of different apparent diffusion coefficient (ADC) of regional of interest (ROI) selection for magnetic resonance diffusion-weighted imaging (DWI) in identifying differentiation degree, lymph nodes metastasis and lymphovascular invasion in preoperative cervical cancer. Methods　Forty patients with cervical cancer confirmed by pathological examination in our hospital were enrolled. All patients underwent MRI routine scanning sequence and DWI sequence using AVANTO 1.5 T magnetic resonance. According to the pathological stage after operation, the degree of tumor differentiation, patients were divided into two groups (high grade and low grade groups). According to the lymph node metastasis patients were divided into two groups (metastasis group and non-metastasis group).According to the lymphovasvular invasion patients were divided into two groups (positive group and negative group).Three methods of ROI measurements (area ROI, volume ROI and small ROIs) were used on DWI images of cervical caner cancer by two physicians with different seniority independently measuring apparent diffusion coefficient (ADC) value of lesions. Differences between groups were analyzed. The parameters with significant differences were analyzed by receiver operating characteristic (ROC) curve and obtaining the area under the curve (AUC), corresponding optimum cutoff value, and diagnostic performance. Results　All the ADC values obtained by three methods of ROI measurements had good repeatability, and the ADC values of volume ROI were the best. There were significant differences in mean value of area ROI and volume ROI between high-middle grade and low grade cervical cancer (P＜0.01). The ROC curve was used to identify high-middle grade and low grade using the ROI volume measurement with the biggest AUC: 0.854, the optimal cutoff value was 0.602×10-3 mm2/s. There were significant differences in mean ADC of volume ROI and the max ADC-minimum ADC of small ROI lymph node metastasis between metastasis group and non-metastasis group (P＜0.05). The ROC curve was used to identify high-middle grade and low grade using the ROI volume measurement with the biggest AUC: 0.788, the optimal cutoff value was 0.594×10-3 mm2/s. The minimum ADC of area ROI, the mean ADC and the minimum ADC of the volume ROI, the mean ADC and the minimum ADC of the small ROC were significantly different in lymphvascular invasion (P＜0.05). The ROC curve was used to identify lymphvascular invasion using the ROI volume measurement with the biggest AUC:0.910, the optimal cutoff value was 0.439×10-3 mm2/s. Conclusion　The ADC measured by volume ROI method is the best repeatability among the three ROI methods. The mean ADC of the volume ROI method and the minimum ADC of the volume ROI method are the best predictive value in the differentiation degree, lymph nodes metastasis and lymphovascular invasion of cervical cancer in turn.

Key words: uterine cervical neoplasms, lymphatic metastasis, diffusion magnetic resonance imaging, ROC curve, area under curve, magnetic resonance apparent diffusion coefficient, differentiated degree, lymphovascular invasion