Application of Self-Organizing Artificial Neural Networks on Simulated Diffusion Tensor Images

Figure 1

Illustration of the training process. (a) Initial random state of the lattice. The input data vector is displayed here as 𝐼(𝑡). Randomly initialized network after a learning step; intermediate stage of self-organization. Best match is assigned as winning node. Updating the weight allows the network to find its best matching nodes in the discrete output space (5). The nodes within the neighborhood ℎ_𝑗 learn from the winning node. (b) Fully trained network after 𝑛 iterations: Structured input space.