Learning Effective Connectivity Network Structure from fMRI Data Based on Artificial Immune Algorithm

The flowchart of the proposed AIAEC algorithm, where an effective connectivity network with the best K2 is obtained by the antibody immune optimization process

The mapping relationship between a brain network and its corresponding candidate solution.

Many approaches have been designed to extract brain effective connectivity from functional magnetic resonance imaging (fMRI) data. Here a new algorithm is developed to infer the effective connectivity between different brain regions by combining artificial immune algorithm (AIA) with the Bayes net method, named as AIAEC. Brain effective connectivity network is mapped onto an antibody, and four immune operators are employed to perform the optimization process of antibodies, including clonal selection operator, crossover operator, mutation operator and suppression operator, and finally gets an antibody with the highest K2 score as the solution. 

Effective connectivity is the influence that one neuronal system exerts over another between brain region. Specifically, effective connectivity can describe the directed networks in the resting state and specific changes of baseline brain activity in some diseases. The effectiveness of AIAEC has been experimentally verified. Moreover, AIAEC is superior to the other existing 10 algorithms in the majority of the datasets. The advantages of AIAEC (e.g., shorter session duration and higher noise-tolerance ability) imply that it is promising for practical applications in the neuroimaging studies of pediatric, geriatric subjects and neurological patients.