Purpose: To investigate the ability of blood oxygen level-dependent (BOLD) MRI to depict clinically significant prostate tumor hypoxia.
Methods and materials: Thirty-three patients with prostate carcinoma undergoing radical prostatectomy were studied preoperatively, using gradient echo sequences without and with contrast medium enhancement, to map relative tissue oxygenation according to relaxivity rates and relative blood volume (rBV). Pimonidazole was administered preoperatively, and whole-mount sections of selected tumor-bearing slices were stained for pimonidazole fixation and tumor and nontumor localization. Histologic and imaging parameters were independently mapped onto patient prostate outlines. Using 5-mm grids, 861 nontumor grid locations were compared with 237 tumor grids (with >50% tumor per location) using contingency table analysis with respect to the ability of imaging to predict pimonidazole staining.
Results: Twenty patients completed the imaging and histologic protocols. Pimonidazole staining was found in 33% of nontumor and in 70% of tumor grids. The sensitivity of the MR relaxivity parameter R(2)* in depicting tumor hypoxia was high (88%), improving with the addition of low rBV information (95%) without changing specificity (36% and 29%, respectively). High R(2)* increased the positive predictive value for hypoxia by 6% (70% to 76%); conversely, low R(2)* decreased the likelihood of hypoxia being present by 26% (70% to 44%) and by 41% (71% to 30%) when combined with rBV information.
Conclusion: R(2)* maps from BOLD-MRI have high sensitivity but low specificity for defining intraprostatic tumor hypoxia. This together with the negative predictive value of 70% when combined with blood volume information makes BOLD-MRI a potential noninvasive technique for mapping prostatic tumor hypoxia.