Quantitative analysis of heterogeneous tumor enhancement pattern and correlation with outcome in cervical cancer

Abstract

5095 Background: Pixel-by-pixel analysis of signal intensity (SI) using dynamic contrast enhanced MRI (DCE- MRI) in cervical cancer has been reported to predict treatment outcome. Low enhancement correlates with treatment failure presumably due to poor blood supply/hypoxia. However, within a heterogeneous tumor, the relative volume and degree of low-enhancement regions that correlate with failure is uncertain. This project analyzed the threshold of low-enhancement pixels within the tumor to optimize the predictive power for treatment outcome. Methods: 101 patients with advanced cervical cancer underwent DCE-MRI early during radiation/chemotherapy (2 weeks after therapy start). The SI of each pixel was calculated from the time-intensity curve of the DCE-MRI at the plateau phase. The SI of all pixels within the tumor was plotted as a pixel SI spectrum. The lowest 2.0th through 40th percentiles of the pixel spectrum were used to define the low-enhancement volume, and each percentile category was correlated with tumor recurrence using logistic regression analysis. Results: The significance levels to predict tumor recurrence with different relative volume thresholds are shown in the Table. Prediction was most significant when thresholds of 3.5th to 7.0th percentiles were used to define the low-enhancement tumor volume. The predictive power decreased when the lowest percentile was defined as >7.5th or <3.0th percentiles of the pixel spectrum. Conclusion: Based on this limited data, the best predictive value is obtained with pixel thresholds at the 3.5th through 7th percentiles of the heterogeneous tumor spectrum. Pixels in this range may reflect tumor cells with poor blood supply/hypoxia and less response to treatment. Higher percentiles may include cell populations with variable blood supply, and lower percentiles than the 3.0th may represent partial volume effects at the tumor margin. These findings require further study. No significant financial relationships to disclose.