[Power Doppler Index for preoperative ovarian tumors discrimination]. / Valeur de l'index de vascularisation (Power Doppler Index) pour la discrimination des masses ovariennes en préopératoire.

ABSTRACT

OBJECTIVE: To assess the performance of a power doppler vascularity index in the preoperative diagnosis of ovarian malignancy. PATIENTS AND METHODS: Two successive series of adnexal masses (N=101 and N=82) were examined prospectively with power Doppler before surgical treatment. A vascularity index, called Power Doppler Index (PDI), based on the number of colored pixels, was estimated on selected frames (defined region of interest covering the entire tumor) of the tumors using an in house color-quantifying program. The first study was monocentric to test the inter- and intraobserver reproducibilities and the availability of the parameter. The second was multicentric to validate the first results. The sensitivity and specificity of PDI, resistance index (RI) and subjective visual scoring were compared using receiver operating characteristic (ROC) curves. RESULTS: Histology identified 23 malignant and 78 benign lesions in the first study and 34 malignant and 48 benign tumors in the second. PDI was considerably higher in malignant than in benign lesions (0.34+/-0.04 vs 0.12+/-0.06 (P<0.001)). Intra-observer and interobserver variability of PDI was low (intraclass correlation coefficients estimated at 0.99 and 0.98, respectively). The PDI cut-off value for differential diagnosis was set at 0.265 (26.5% of the tumor is colored) in the first study and 0.107 in the second. Sensitivity and specificity were 100% (95% CI [87.8; 100.0]) and 97.4% (95% CI [91.0; 99.7]), respectively, for PDI in the first study compared to 75% (95% CI [72.9; 90.7]) and 60% (95% CI [56.3; 92.5]) in the second study. Using logistic regression, visual Doppler scoring performed best. DISCUSSION AND CONCLUSION: The Power Doppler Index obtained using our color quantifying software has discriminating power for adnexal masses but visual Doppler scoring performs best.