Comparison of functional localization accuracy with different co‐registration strategies in presurgical fMRI for brain tumor patients1. Issue 7 (13th June 2018)

• Record Type: Journal Article
• Authors: Jen, Mu‐Lan; Hassan, Islam; Hou, Ping; Li, Guang; Kumar, Ashok J.; Prabhu, Sujit S.; Colen, Rivka R.; Liu, Ho‐Ling
• Is Part Of: Medical physics. Volume 45:Issue 7(2018)
• Page Start: 3223
• Page End: 3228
• Publication Date: 2018-06-13
• Subjects: brain tumors -- functional MRI (fMRI) -- image registration -- presurgical; Medical physics -- Periodicals; Medical physics; Geneeskunde; Natuurkunde; Toepassingen; Biophysics; Periodicals; Periodicals; Electronic journals; 610.153
• DOI: 10.1002/mp.12999
• Languages: English
• ISSNs: 0094-2405

Abstract

Abstract : Purpose: Presurgical fMRI is an important tool for surgery navigation in achieving maximum resection of a brain tumor. However, the functional localization accuracy may be compromised by spatial transformation from echo‐planar images to high‐resolution structural images. We evaluated functional localization errors associated with the spatial transformation process using three algorithms commonly applied to the presurgical fMRI in the clinic. Methods: MR images of 20 brain tumor patients for presurgical evaluation of eloquent areas near motor cortices were analyzed. All fMRI data were spatially transferred to 3D T1‐weighted images using three algorithms: (a) coordinate matching (CM), (b) automated registration (AR), and (c) AR plus manual adjustment (ARadj ). Activation clusters overlaid on original echo‐planar images were manually delineated on slice‐matched 2D T1‐ weighted images and then transferred to the 3D T1‐weighted image volume, and served as the reference localization. Functional localization errors were estimated by measuring the distance between the reference localization and the activation cluster after spatial transformation and then compared for the three algorithms. Results: The 3D Euclidean distance for AR (10.2 ± 4.9 mm) was found to be significantly larger ( P < 0.05) than those for CM (5.6 ± 2.6 mm) and ARadj (5.8 ± 3.0 mm) algorithms. The difference between the localization errors in CM and ARadj was not statistically significant. Conclusions: A procedure was proposed to evaluate functional localization errors associated with spatial transformation in presurgical fMRI. Our results highlighted the necessity of routine quality control for the AR processing in the clinic.