Image registration and quantitative analysis

The whole-brain imaging data were registered to a 3D reference brain atlas called the Allen Common Coordinate Framework (CCF)⁶⁰. The imaging data recorded from the PI channel were downsampled to a spatial resolution of 10 × 10 × 10 μm³. First, the imaged brains were corrected with rigid registration. A greyscale-based 3D affine registration was used for alignment to the Allen Brain Atlas, after which dense landmark-based 2D registration was performed in local regions. All registration processes were performed using Elastix⁶¹. Subsequently, the traced points of neuron reconstructions were transformed into Allen Brain Atlas space using the parameters for the registration transformation. Two experienced analysts checked the image registration results by back-to-back manual confirmation. Morphological features were quantitatively analyzed using Amira and custom-written software in MATLAB. The branch lengths for both axons and dendrites were measured by summing the distances from the traced points to their parent nodes. The branch numbers for both axons and dendrites were measured by counting the segments that connected to the parent nodes.