Pitfalls of Thresholding Statistical Maps in Presurgical fMRI Mapping

Krzysztof J. Gorgolewski, Mark Bastin, Laura Rigolo, H. A. Soleiman, Cyril Pernet, Amos Storkey, Alexandra Golby

Research output: Contribution to conferencePosterpeer-review

Abstract

There is significant variability in the selection of methods used to threshold fMRI activation maps acquired for brain tumour presurgical planning. In a review of 50 recent papers, only 12% of studies claimed to use some kind of Family Wise Error (FWE) correction for multiple comparison testing. 42% of these studies used a p-value threshold lower than the standard 0.05 in an attempt to minimize the number of false positives, while 22% did not use the same threshold for all of the subjects, with the threshold being manually adjusted on per subject basis. What is more, most of the studies used simple thresholds without taking the spatial properties of the statistical maps into account; only 4% used cluster size as an additional threshold. Although thresholding methods used in neuroscience fMRI studies have greatly improved in the last few years, there still remains a lack of consensus in the clinical literature about how to identify activation boundaries accurately and objectively. As a first step towards developing robust frameworks for assessing thresholding methods for tumour resection, we investigated how several automated thresholding methods affect the distance between the activation areas determined from fMRI experiments and the tumour boundary defined on structural MRI, and how this data might potentially change surgical practice.
Original languageEnglish
Publication statusPublished - 2011
EventISMRM 19th Annual Meeting & Exhibition - Montreal, Quebec, Canada
Duration: 7 May 201113 May 2011

Conference

ConferenceISMRM 19th Annual Meeting & Exhibition
Country/TerritoryCanada
CityMontreal, Quebec
Period7/05/1113/05/11

Fingerprint

Dive into the research topics of 'Pitfalls of Thresholding Statistical Maps in Presurgical fMRI Mapping'. Together they form a unique fingerprint.

Cite this