Brain diffusivity pattern is individual-specific information

Abstract

The human brain is composed of complex networks of 100 billion neurons that underlie its higher functions. The set of neural connections in the brain has recently attracted growing interest from the scientific community. It is important to identify individual differences in these neural connections to study the background of individual differences in brain function and performance. In the present study, we investigated whether the pattern of brain diffusion, reflecting neural connections, is discernibly different among individuals; i.e., whether brain diffusivity is personally identifiable information. Using diffusion tensor imaging data from 224 healthy subjects scanned twice at an interval of about 1year, we performed brain recognition by spatial normalization of fractional anisotropy maps, feature extraction based on Principal Component Analysis, and calculation of the Euclidean distances between image pairs projected into the subspace. Even with only 16 dimensions used for projection, the rank-one identification rate was 99.1%. The rank-one identification rate was 100% with ⩾32 dimensions used for projection. The genuine accept rates were 95.1% and 100% at a false accept rate of 0.001%, with 16 and ⩾32 dimensions used for projection, respectively. There were no large differences in the Euclidean distance among different combinations of scanners used or between image pairs with and without scanner upgrade. The results indicate that brain diffusivity can identify a specific individual; i.e., the pattern of brain diffusion is personally identifiable information. Individual differences in brain diffusivity will form the basis of individual differences in personality and brain function.