Objectives: To analyse the long-term feasibility and limitations of presurgical fMRI in a cohort of tumour and epilepsy patients with different MR-scanners at 1.5 and 3.0 T.
Methods: Four hundred and ninety-one consecutive patients undergoing presurgical fMRI between 2000 and 2012 on five different MR-scanners using established paradigms and semi-automated data processing were included. Success rates of task performance and BOLD-activation were determined for motor and somatosensory somatotopic mapping and language localisation. Procedural success, failures and imaging artifacts were analysed. MR-field strengths were compared.
Results: Two thousand three hundred fifteen of 2348 (98.6 %) attempted paradigms (1033 motor, 1220 speech, 95 somatosensory) were successfully performed. 100 paradigms (4.3 %) were repetition runs. 23 speech, 6 motor and 2 sensory paradigms failed for non-compliance and technical issues. Most language paradigm failures were noted in overt sentence generation. Average significant BOLD-activation was higher for motor than language paradigms (95.8 vs. 81.6 %). Most language paradigms showed significantly higher activation rates at 3 T compared to 1.5 T, whereas no significant difference was found for motor paradigms.
Conclusions: fMRI proved very robust for the presurgical localisation of the different motor and somatosensory body representations, as well as Broca's and Wernicke's language areas across different MR-scanners at 1.5 and 3.0 T over 13 years.
Key points: • Standardised presurgical motor and language fMRI is robust across various MRI platforms. • Motor fMRI is less dependent on field strength than language fMRI. • fMRI task failures are relatively low and are reduced by paradigm repetition.
Keywords: Brain tumour; Functional MRI; Language; Motor; Somatosensory.