Room: AAPM ePoster Library
Purpose: Brain shift during neurosurgical tumor resection affects the accuracy of image-guidance and the localization of residual tumor and important anatomical structures. In this study, the potential of viscoelastic finite element biomechanical modelling (FEM) was investigated to predict the localization of residual tumor in intraoperative images.
Methods: Intraoperative Magnetic Resonance (iMR) images after partial tumor resection were retrospectively obtained for five patients under an IRB approved study. For each patient, an individualized viscoelastic FEM (Zener Model) was constructed with brain tissues semi-automatically segmented on both preoperative MR (pMR) and iMR (cortex, cerebellum, head, craniotomy, meninges, tumor). Assuming gravity-induced brain shift and complete cerebrospinal fluid (CSF) drainage, brain stiffness was optimized based on the Dice Similarity Coefficient (DSC) between the smoothed cortical surface from the iMR and the predicted displaced surface from the pMR. Based on the final brain displacement predictions, the Student’s t-test was used to test for significant improvements in the false negative (FNF) and true positive fractions (TPF) of the residual tumor volume using the FEM compared to standard rigid registration.
Results: The optimal brain shear modulus for each patient ranged 0.6 to 1.4 kPa, with a median value of 1.12 kPa. The average FNF was 0.53, range 0.27-0.79 and 0.33, range 0.16-0.55 for the rigid registration and FEM-based registration, respectively, (P=0.005). The average TPF was 0.43, range 0.18-0.65 and 0.62, range 0.47-0.78, for rigid registration and FEM-based registration, respectively (p<0.001).
Conclusion: This pilot study evaluates the accuracy of gravity-induced viscoelastic brain shift FEM to predict tumor displacement demonstrating a significant improvement over rigid registration, currently available. Future work will investigate the impact of additional brain shift forcing conditions (swelling, shrinking, partial CSF drainage, tumor resection) on the accuracy metrics to further improve the TPF and FNF.
Funding Support, Disclosures, and Conflict of Interest: This work was supported through the Helen Black Image-Guided Fund and a donation from the Apache Corporation.