Accelerated Clustered Sparse Acquisition to Improve Functional MRI for Mapping Language Functions

Abstract

Abstract Background Functional magnetic resonance imaging (fMRI) is a useful method for noninvasive presurgical functional mapping. However, the scanner environment is inherently unsuitable for the examination of auditory and language functions, due to the loud acoustic noise produced by the scanner. Interleaved acquisition methods alleviate this problem by providing a silent period for stimulus presentation and/or response control (sparse sampling) but at the expense of a diminished amount of data collected. There are possible improvements to these sparse acquisition methods that increase the amount of data by acquiring several images per event (clustered sampling). We tested accelerated clustered fMRI acquisition in comparison with conventional sparse sampling in a pilot study. Methods The clustered and sparse acquisition techniques (7.4 minutes scanning time per protocol) were directly compared in 15 healthy subjects (8 men; mean age: 24 ± 3 years) using both a motor (tongue movement) and a language (overt picture-naming) task. Functional imaging data were analyzed using Statistical Parametric Mapping software (SPM12 Wellcome Department of Imaging Neuroscience, London, UK). For both tasks, activation levels were compared and Euclidean distances (EDs) between cluster centers (i.e., local activation maxima and centers of gravity) were calculated. Overlaps and laterality indices were computed for the picture-naming task. In addition, the feasibility of the clustered acquisition protocol in a clinical setting was assessed in one pilot patient. Results For both tasks, activation levels were higher using the clustered acquisition protocol, reflected by bigger cluster sizes (p < 0.05). Mean ED between cluster centers ranged between 9.9 ± 5.4 mm (left superior temporal gyrus; centers of gravity) and 16.6 ± 13.2 mm (left inferior frontal gyrus; local activation maxima) for the picture-naming task. Overlaps between sparse and clustered acquisition reached 88% (Simpson overlap coefficient). A similar activation pattern for both acquisition methods was also confirmed in the clinical case. Conclusion Despite some drawbacks inherent to the acquisition technique, the clustered sparse sampling protocol showed increased sensitivity for activation in language-related cortical regions with short scanning times. Such scanning techniques may be particularly advantageous for investigating patients with contraindications for long scans (e.g., reduced attention span).