SOM Based Diffusion Tensor MR Analysis

Self-organizing map (SOM) is an unsupervised learning method which is used in training neural networks. It is considered to be unsupervised because the correct output, which corresponds to the input data, is not specified. Therefore SOM classifies data into groups. In our study, the classification is actually the tracking of eigenvectors defining the principal diffusivity of the fibers in the diffusion tensor magnetic resonance imaging (DTMRI). SOM is first implemented on synthetic diffusivity paths, which is achieved easily by assigning random eigenvectors to a region of interest simulating the imaging matrix. Following the verification of the SOM application on synthetic data, the implementation follows on diffusion weighted human brain images. The idea of the proposed method is to accomplish the fiber pathway based on learning rule by starting from a single, selected node taking every node into account. Implementing SOM on DTMRI tractography was successful and future work relies in 3D diffusion tensor tractography based on SOM.