Regional 18F-fluorodeoxyglucose (FDG) uptake in the brain on whole body PET-CT of breast cancer patients receiving chemotherapy.

Objectives The ability to analyze large numbers of clinically-acquired oncologic PET-CT datasets may lead to an improved understanding of the relationship between various therapeutic agents and clinical phenomenon such as chemotherapy-related cognitive dysfunction. This exploratory study analyzed relative regional 18F-FDG uptake in the brain of breast cancer patients on PET-CT following various chemotherapy regimens.

Methods A pipeline was constructed to continuously and automatically receive, de-identify, and process clinically acquired whole body (WB) PET-CT DICOM datasets from a PACS database. Processing included evaluation of various parameters (patient direction, presence of i.v. contrast, arm position), identification of the skull vertex, and extraction of cortical activity levels in sub-regions of the brain. Regions of interest (ROIs) were defined by atlas based segmentation following multi-resolution coregistration of a T1-weighted MRI brain template with PET-CT via combined affine and b-spline transformations in a multi-resolution registration. Normalized mutual information was used as the metric in the registration. A label image (transformed along with the template) was used to segment (left (l) and right (r)) frontal (F), parietal (P), temporal (T), frontal cingulate (FC), and parietal cingulate (PC) regions from the PET image, excluding regions of potential infarct or atrophy. Non-cortical PET voxels were also excluded using an intensity-based approach. All coding was performed in python, and the registration done via calls to elastix. [1,2] Assessment of the segmentation quality was performed on a subset of 64 processed images by visually assessing alpha blended overlays of transformed label maps over 3 axial brain slices in the PET images. Mean cortical activity levels (Bq/cm3) of the ROIs were normalized by mean ipsilateral cerebellar activity. Two-sample t-tests were used to evaluate for significant differences in the mean values of normalized regional activity (NrA).

Results A total of 104 PET-CT exams of 52 patients were analyzed, requiring an average of 3 minutes per exam to complete all analysis steps. Alignment and segmentation was acceptable in 98.4% of cases. Mean clinical follow-up after PET was 4.3 months (range 1-9.9). The change in NrA across all regions on PET before and after chemotherapy administration was &#8804; 1.5%. Mean % change in (l)T uptake for patients receiving chemotherapy between PET studies (Pchemo) was 1.37% vs. -0.69% for patients receiving no therapy or biologic therapies only (Pnone) (p < .05). Mean % change in (l)F uptake for Pchemo was 1.25% vs. -0.76% for Pnone (p < .05). A non-significant trend of increased uptake in (l)P was observed for Pchemo (1.50%) vs. Pnone (-0.12%). There was no significant difference in the level of 18F-FDG NrA in patients who had previously received chemotherapy at any point in their management versus those who had not. NrA in the (r)FC, bilateral F, C, P, and T regions was significantly decreased in patients with symptoms of anxiety, as documented in the medical record following chemotherapy, compared to asymptomatic patients (p<.05-.01).

Conclusions Though exploratory in nature, this study demonstrates the feasibility of rapidly analyzing regional metabolism in large numbers of oncologic PET-CT datasets using pipelines that accurately segment and quantify activity. Using this methodology, significant differences in regional brain metabolism were identified between asymptomatic patients and those with anxiety. Minimal changes in regional activity were measured during chemotherapy, in agreement with previously published studies of other cancer populations. [3,4] Scaling of these techniques will be leveraged in future analyses of larger sets of WB PET studies to evaluate potential effects of specific therapies on regional brain metabolism.