TARGET AUDIENCE – Clinical radiologists, radiotherapists, and scientists involved in the diagnosis and treatment of cervical cancer.Previous studies have demonstrated that MR imaging texture analysis varies over the course of therapy in a wide range of malignancies, including breast cancer (1), osteosarcoma (2), and rectal cancer (3). Current non-surgical treatment for cervical tumors involves external beam radiotherapy (EBRT) to the pelvis with concurrent cisplatin based chemotherapy, followed by intracavitary high dose rate (HDR) brachytherapy. The goal of our study is to examine the effects of chemoradiation on cervical tumor blood oxygen level dependent (BOLD) MRI texture.This retrospective, HIPAA-compliant study was approved by our institutional human subjects review committee. 25 patients (35 – 66 years; 49 +/- 10.2 years) with varying stages of cervical cancer (from IB to IIIB) were imaged with BOLD MRI at 3 time-points: before treatment (time point 1), after treatment with chemotherapy and EBRT (time point 2), and after completion of the EBRT and HDR brachytherapy. (time point 3). Subjects were imaged with either a 1.5 or 3T MR scanner (Signa Excite HD, GE Healthcare, Waukesha, WI, USA) and either an 8 or 32-channel body coil. BOLD MRI parameters were: TR/TE/flip angle = 87ms/7-42ms/40°, FOV = 32-34cm, slice thickness 4.0 mm, skip 0.5 mm, and 256x256 matrix. R2* maps were generated and imported into a separate computer workstation for texture analysis. A volume of interest (VOI) was drawn, encompassing the entire cervical tumor, or if no tumor was visible, then the entire cervix. VOI’s were imported into a custom designed MatLab program, which quantified different measures of image heterogeneity, including skewness, kurtosis, entropy, and area under the curve (AUC) using the cumulative volume histogram of pixel distribution. Statistical significance between groups was compared with a Student’s T-test. Over the course of treatment (comparing time point 1 with time point 3), there is a significant decrease in the AUC (from 69.4% to 58.8 P<0.01). Furthermore, there are significant decreases in skewness (from 0.8 to 0.3, P<.01) and kurtosis (from 4.3 to 3.1, P<.01) (figure 2). Out of 22 tumors, 40% (8/20) recurred. Tumors which recurred tended towards a higher AUC at time point 2 (AUC = 61.8% vs. 55.6%; p=0.08). Tumors that did not recur tended towards a lower skewness of their pixel distribution, 0.09 vs. 0.47 for recurrent tumors (P=0.06) at time point 2. There was no significant difference in other measures of heterogeneity (kurtosis and entropy) between the tumors that recurred and those that did not.In our study, after initial treatment (time point 2), tumors that recurred tended to have a higher AUC, and tended to have a positive skew in their image histogram. As tumors undergo treatment, the AUC and skewness decreases significantly. Our findings indicate that BOLD MRI texture analysis can be used to assess long-term response to therapy after initial treatment and to follow tumors during treatment. Further studies using BOLD MRI texture analysis in cervical cancer may help elucidate the utility of this technique in the course of treatment of women with cervical cancer.