The Parametric Response Map (PRM) compares spatially paired voxels of co‑registered images at different time‑points to quantitatively analyze significant changes in the histology of the tumor over time, which provided an opportunity to quantify changes in tissue properties in response to therapy. Studies show that the PRM metric applied to apparent diffusion coefficient (ADC) maps are predictive of clinical progression for head and neck squamous cell carcinoma ¹ and one-year survival of glioma patients ^2–5, whereas percentage‑change in mean ADC failed to predict outcome. Thus, temporal dynamics within heterogeneous tumors are often lost in scalar metrics ⁶. The PRM method has been extended across modalities and serves as a generalizable approach for quantifying and spatially visualizing tissue changes longitudinally ^7,8. The purpose of this study was to investigate the possibility of adapting the PRM method for use with normalized axial fluid‑attenuated inversion recovery (FLAIR) images (PRM_rFLAIR) to predict overall survival (OS) and time to progression (TTP), thus functioning as an early indicator of tumor recurrence.Patients (n=52) with grade 3/4 glioma were recruited for a prospective single-center clinical trial. Patients were separated into groups; those surviving less than one year (non-responders) and those surviving at least one year (responders) after treatment. Each patient was treated with a total radiation dose of 60‑70 Gy over 30 fractions. Radiotherapy is typically combined with concurrent TMZ, followed by 6 months of adjuvant chemotherapy. FLAIR images were acquired post-surgical resection, prior to treatment (0‑week image) and 10 weeks after commencement of treatment. MRI parameters can be found in Figure 5. The 10‑week FLAIR images were then co‑registered to the 0-week images using Elastix ⁹ (Figure 1). For both time‑points, volumes of interest (VOI) were used to outline the tumor volume. Both FLAIR images were normalized using the cerebellum to obtain a relative FLAIR (rFLAIR) image, ensuring signal intensities were comparable between time‑points. Voxels present in the union of the two VOIs for the rFLAIR images were used to evaluate the PRM_rFLAIR values. The calculated PRM_rFLAIR values were categorized into three groups based on significant changes in signal intensity: decrease (PRM_rFLAIR‑, blue voxels), increase (PRM_rFLAIR+, red voxels) and unchanged (PRM_rFLAIR0, green voxels) (Figure 1). Significant changes were determined using a 95% CI threshold calculated from a linear least squares regression over contralateral healthy brain tissue for ten randomly selected patients. The mean voxel intensity within the tumor VOIs were calculated for both 0 and 10‑week rFLAIR images. Patients were categorized into the PRM_rFLAIR groups (red, blue and green voxels) using receiver operating characteristic curve analysis to calculate the optimal signal cut‑off (Figure 3). Within each cut‑off, a Kaplan‑Meier estimate was used to evaluate the survival probability, after treatment, of both responders and non‑responders and significance was tested using a Log Rank (Mantel‑Cox) test.The optimal cut‑off for voxel signal intensities for both the PRM_rFLAIR+ and PRM_rFLAIR‑ to determine one-year survival is shown in Figure 3. From this cut-off, patients were characterized as either responders or non‑responders. The categorized PRM_rFLAIR patient samples were analyzed using the Kaplan‑Meier tumor progression curve (Figure 4). Figure 4 reveals the metric PRM_rFLAIR+ allows for predicting mean and median TTP for non‑survivors which was significantly less (p < 0.001) than the estimated mean and median TTP for survivors. OS was also analyzed using Kaplan‑Meier estimation. Using the metric PRM_rFLAIR+, the mean and median survival time for non‑responders was estimated to be significantly less (p = 0.02) than for responders. Similarly, by using PRM_rFLAIR-, the predicted mean and median survival time for non-responders was also significantly less (p = 0.008) than the mean and median survival time for responders.Results show patients who had a significantly increased change in voxel signal intensity (PRM_rFLAIR+) were classified as non-responders with significantly shorter OS and TTP versus responding patients with insufficient PRM_rFLAIR+ signal. Additionally, patients with a significantly decreased change in voxel intensity (PRM_rFLAIR-), had a significantly longer OS. This study shows that predicting TTP and OS is achievable using PRM with rFLAIR maps for glioma patients treated with TMZ/IR.FLAIR images are routinely acquired for glioma patient clinical diagnosis and management. We demonstrated that these FLAIR images could be normalized to generate maps that could be used to provide quantitative PRM maps. This study provides the first demonstration of spatially following and quantifying FLAIR signals in patients over time and opens up new possible diagnostic uses for PRM_rFLAIR in the clinic.