Abstracts

Characterizing brain perfusion inter-individual functional variability in SPECT is a key issue to better understand the physiology and physio-pathology in order to model any perfusion pattern and to support diagnosis. Volume of Interest (VOI)-based anatomic standardization analysis has been widely used to study normal perfusion patterns in SPECT ¹. Principal component analysis has also been proposed to study functional variability in healthy subjects ². The purpose of this study was to explore the structure of statistical dependencies, inherent to a perfusion pattern as observed in a population of SPECT data, using correspondence analysis (CA). We studied two perfusion patterns: (1) normal perfusion in healthy subjects and (2) ictal perfusion in mesial temporal lobe epilepsy (MTLE) patients.
SPECT scans from a group of 27 healthy subjects (12 men and 15 women, aged 20 to 56 years) were used to study normal perfusion. We selected 10 operated seizure-free MTLE patients (six men and four women, aged 19 to 43 years), who showed a typical MTLE ictal SPECT pattern ³ during pre-surgical investigation. VOIs specifically designed to study MTLE were used to automatically perform perfusion measurements, using a previously proposed anatomic standardization method ⁴. For each population, we then used CA and hierarchical clustering ⁵ to explore statistical dependencies between the shapes of perfusion measurements distributions within each VOI.
For the normal group (1), the first three principal components were selected in order to recover 49% of global inertia, highlighting three groups of anatomical structures: internal structures (caudate nucleus, thalamus, lenticular nucleus), mesio-temporal structures (amygdala, hippocampus) and remaining cortex. Within each of these groups, perfusion measurements may then be considered as statistically dependent.
For the patients group (2), the first two principal components were considered sufficient to extract main information (70.9 % of global inertia). Similarly to healthy subjects, the analysis highlighted three groups corresponding to internal structures, temporal structures and remaining cortex. Nevertheless, the temporal pole and the lenticular nucleus, respectively classified in the remaining cortex group and in the internal structures group within the healthy subjects analysis, appeared now as most relevant structures of the temporal structures group.
These results revealed a structure of dependencies between perfusion measurements, highlighting the relevant role of the temporal pole and the lenticular nucleus in ictal perfusion. This suggests that CA is a promising approach to model functional variability among a population of SPECT data.
¹Van Laere K [italic]et al[/italic] 2001 [italic]EJNM[/italic] 873-887
²Pagani M [italic]et al[/italic] 2002 [italic]EJNM[/italic] 67-75
³Ho S [italic]et al[/italic] 1996 [italic]Epilepsia[/italic] 788-795
⁴Grova C [italic]et al[/italic] 2002 [italic]CARS conference[/italic], 450-455
⁵Lebart L [italic]et al[/italic] 1984 [italic]Wiley series in probability and math. statistics[/italic], NY