Adaptive neuro-fuzzy inference system-based chaotic swarm intelligence hybrid model for recognition of mild cognitive impairment from resting-state fMRI

Abstract

Individuals diagnosed with mild cognitive impairment (MCI) are at high risk of transition to Alzheimer’s disease (AD). Resting-state functional magnetic resonance imaging (rs-fMRI) is a promising neuroimaging technique for identifying patients with MCI. In this paper, a new hybrid model based on Chaotic Binary Grey Wolf Optimization Algorithm (CBGWO) and Adaptive Neuro-fuzzy Inference System (ANFIS) is proposed; namely (CBGWO-ANFIS) to diagnose the MCI. The proposed model is applied on real dataset recorded by ourselves and the process of diagnosis is comprised of five main phases. Firstly, the fMRI data are preprocessed by sequence of steps to enhance data quality. Secondly, features are extracted by localizing 160 regions of interests (ROIs) from the whole-brain by overlapping the Dosenbach mask, and then fractional amplitude of low-frequency fluctuation (fALFF) of the signals inside ROIs is estimated and used to represent local features. Thirdly, feature selection based non-linear GWO, chaotic map and naive Bayes (NB) are used to determine the significant ROIs. The NB criterion is used as a part of the kernel function in the GWO. CBGWO attempts to reduce the whole feature set without loss of significant information to the prediction process. Chebyshev map is used to estimate and tune GWO parameters. Fourthly, an ANFIS method is utilized to diagnose MCI. Fifthly, the performance is evaluated using different statistical measures and compared with different met-heuristic algorithms. The overall results indicate that the proposed model shows better performance, lower error, higher convergence speed and shorter execution time with accuracy reached to 86%.