Figure 2

3D image preprocessing steps improving brain registration. (A) Brain sample pre-alignment via rotation and scaling using Principal Component Analysis (PCA). (B) Cross-section of the binary mask of the brain sample. (C) Contours of the brain regions extracted with Laplacian of Gaussian (LoG) filter. (D) Filtered image; a weighted sum of (B) and (C). (E–H) Registration of two brains taken at P5 and P6 using individual image preprocessing methods as indicated. Registration of raw images (E) results in slight mismatches between Rostral Migratory Streams (RMSs). Registration of binary masks (F) results in a mismatch of the inner structure. Registration of brain region contours (G) may not converge and results in large mismatches between samples. The combination of binary mask and contours (H) solves the problems of (E–G) yielding a higher quality registration.