Figure 1 | Scientific Reports

Figure 1

From: Comparison of the diagnostic accuracy of resting-state fMRI driven machine learning algorithms in the detection of mild cognitive impairment

Figure 1

Pipeline of the procedure: full analysis procedure of the feature extraction, feature selection and classification based on rs-fMRI data. CONN Toolbox was used to analyse rs-fMRI data from 78 participants (n = 32 MCI; n = 46 HC). Preprocessing, feature extraction of three rs-fMRI metrics (ICC, LCOR, fALFF) was done. Then we calculated the average of each metric for different brain regions (132 regions based on ALL and Harvard–Oxford atlas). The extracted parameters (3 × 132 brain regions) were subsequently given to one of the feature selection methods to determine the best subset of ROIs for classification. Five feature selection methods were analysed consisting of 4 optimization and one conventional feature selection algorithm. Two classification algorithms were utilized (SVM, RF) with the algorithms. EA evolutionary algorithm, SFFS sequential floating forward selection, SVM support vector machine, ICC intrinsic connectivity, LCOR local correlation, fALFF fractional amplitude of low frequency fluctuations.

Back to article page