Meeting Banner
Abstract #1809

Co-Registration of MRI Via a Learning Based Fiducial-Driven Registration (LeFiR) Scheme: Evaluating Laser Irradiation Changes for Glioblastomas and Epilepsy

Tao Wan1, B.Nicolas Bloch2, Shabbar Danish3, Anant Madabhushi1

1Case Western Reserve University, Cleveland, OH, United States; 2Boston University School of Medicine, Boston, MA, United States; 3University of Medicine and Dentistry, New Jersey, New Brunswick, NJ, United States

The purpose of this work is to co-register pre- and post-ablation MRI images for laser-induced interstitial thermal therapy (LITT) of neurological disorders, in order to build an improved model for therapy planning and evaluating imaging related treatment changes in terms of MRI markers for LITT. Despite being a promising treatment option for multiple brain diseases, the effect of LITT on the focal site is currently unknown. The objective of this work is to develop a learning based fiducial driven registration method (LeFiR) for accurately registering pre- and post-LITT brain images. The localized nature of deformation induced by LITT can be well captured and precisely aligned to pre-LITT image via a supervised learning scheme. The identified optimal landmark set is utilized to drive a thin-plate spline (TPS) image registration. The LeFiR method was performed to register pre- and post-LITT brain MR images for treating glioblastoma multiforme (GBM) and epilepsy with LITT. By using LeFiR, more accurate and robust registration results can be achieved according to two registration experiments performed on brain MRI for GBM and epilepsy. The local deformation induced by the LITT procedure can be better captured and recovered by the identified landmark fiducials than simply uniformly picking landmarks. Given that only spatial information is used to determine landmark locations, the LeFiR method has the potential to be adopted in various clinical applications for the purpose of registering different image modalities.