Abstract
1861
Objectives A normal functional imaging brain scan database comprised of multiple normal individuals can be beneficial for statistical analysis. Most analytical techniques create an average mapping of subjects to a preformed atlas, then perform statistical comparison of a subject to the average composite. Normal variation in brain metabolism is natural, especially in the aging population. Comparison of a single subject with the average composite in theory may result in variations greater than 2 standard deviations (SD) for any voxel. This standard deviation may be interpreted as an abnormal finding, suggesting brain injury. The purpose of this IRB approved study was to statistically determine the extent of normal variation in the human brain using an accepted quantitative software based analytical approach.
Methods Normal FDG-PET brain scans were obtained from the Alzheimer’s Disease Neuroimaging Initiative2 (ADNI). These subjects were declared cognitive normal by the ADNI team and showed no signs of depression, mild cognitive impairment, or dementia. FDG-PET brain scans were acquired at an ADNI approved facility using a standardized protocol. 26 clinically normal individual scans were selected at random with an average age of 75. All subjects had a mini-mental state examination (MMSE) greater than 27 (average 29). A normal database was created using an established quantitative software approach. Each individual normal scan was compared to the normal ADNI database, removing that individual from the database. Computer maps were created examining the standard deviations (SD) from 2 - 7 above and below the average voxel for each subject.
Results When examining hypometabolism in subjects, there was a false positive rate of 100% at 2 standard deviations. Findings of focal areas of relative reduction diminished between 4 and 5 deviations from the norm as expected. Relative increases using quantitative analysis displayed a false positive rate of 92% up to 4 standard deviations and diminished to 8% at a standard deviation of 6. No normal subjects displayed a recognizable pattern for any known dementia or major vascular event.
Conclusions Discussion: More than statistically expected false positive findings may be due to combination of non-normal distribution of data points and limited number of subjects in the normal database. Caution must be followed in quantitative interpretation of functional brain imaging studies when no recognizable pattern of disease is present. Conclusions: It would be important to consider the amount of variations in normals while performing statistical analysis of functional brain imaging. Quantitative analysis is helpful in identifying recognizable patterns of cerebral dysfunction. Although caution should be exercised when making a diagnosis based upon random focal areas of either relative hypometabolism or hypermetabolism which fail to exceed normal variation.