Pixel analysis of MR perfusion imaging in predicting radiation therapy outcome in cervical cancer

The purpose of this study was to assess heterogeneity of tumor microcircula tion determined by dynamic contrast-enhanced magnetic resonance (MR) imagin g and its prognostic value for tumor radiosensitivity and long-term tumor c ontrol using pixel-by-pixel analysis of the dynamic contrast enhancement. S ixteen patients with advanced cervical cancer were examined with dynamic co ntrast-enhanced MR imaging at the time of radiation therapy. Pixel-by-pixel statistical analysis of the ratio of post- to precontrast relative signal intensity (RSI) values in the tumor region was performed to generate pixel RSI distributions of dynamic enhancement patterns. Histogram parameters wer e correlated with subsequent tumor control based on long-term cancer follow -up (median follow-up 4.8 years; range 3.8-5.2 years). The RSI distribution histograms showed a wide spectrum of heterogeneity in the dynamic enhancem ent pattern within the tumor. The quantity of low-enhancement regions (10th percentile RSI < 2.5) significantly predicted subsequent tumor recurrence (88% vs. 0%, P = 0.0004). Discriminant analysis based on both 10th percenti le RSI and pixel number (reflective of tumor size) further improved the pre diction rate (100% correct prediction of subsequent tumor control vs. recur rence). These preliminary results suggest that quantification of the extent of poor vascularity regions within the tumor may be useful in predicting l ong-term tumor control and treatment outcome in cervical cancer. (C) 2000 W iley-Liss, Inc.