Abstracts

A functional magnetic resonance imaging (fMRI) memory paradigm that will reliably activate medial temporal-lobe structures has an immense potential usefulness in evaluation of patients with temporal-lobe epilepsy. It has been notoriously difficult to activate the hippocampus, but recently some successful paradigms have been devised. To be clinically useful, an fMRI memory test should activate medial structures of both hemispheres reliably in healthy individuals, so that unilateral activation would be clearly abnormal. It should also yield stable results within individuals, so that if changes are observed on retest they can be interpreted as reflecting real changes in brain function. We examined the reliability for site and extent of activations when subjects are retested, using a memory encoding paradigm that has elicited bilateral activation of medial temporal-lobe (MTL) structures in most healthy individuals tested. Seven healthy volunteers and one patient were tested twice on a picture-encoding task during fMRI. Subjects viewed 120 pictures for later recognition, in a blocked design. The two tests used different sets of pictures equal in difficulty; order was counterbalanced. Intervals between tests varied from 2.5 to 20 months (mode=5.5). Stability of memory test scores as well as site and degree of brain activations were analyzed. Memory scores improved in second testing (t=3.91, p[lt].01) and nobody earned a lower score on retest. Three subjects activated essentially the same MTL regions in both sessions, but differences were observed in four: most were in the direction of fewer MTL activations upon retest, but changes were also seen in site and side. These changes were not related to time interval between tests. In contrast, frontal-lobe activations did not disappear on retest for 6 of 7 subjects and indeed tended to increase. The patient, whose second test followed a left anterior temporal-lobe resection and whose memory scores were excellent both times, showed bilateral MTL activations before surgery and none on the left postoperatively. Results were variable among seven healthy subjects retested so far, with some showing similar activations upon retest and others not. No clear pattern emerges from this limited sample, and we are continuing to collect subjects and to investigate possible variables that may contribute to differences when they occur. If these findings remain in a bigger sample, it appears that fMRI activations are a snapshot at a moment in time and may not reflect a basic and enduring status of an individual[rsquo]s memory system, even for a same test paradigm. As habituation plays a role in reducing activations for repeated tests, it may be wise to test twice (using different memory tasks) in both test and retest, to obtain more representative samples. These results suggest that data from retest after surgery or other interventions should be interpreted with caution. (Supported by Canadian Institutes of Health Research)