Diploid Genetic Algorithm-Based Pipeline for the Conversion Prediction of Mild Cognitive Impairment

Mild Cognitive Impairment (MCI) can be considered a formative stage of dementia and its diagnosis may help the clinicians to impediment its headway. This work classifies MCI- Converts and Non-Converts using the structural Magnetic Resonance Imaging data obtained from the Alzheimer's Disease Neuroimaging Initiative. This work investigates the applicability of six feature extraction methods and their permutations, followed by three feature selection methods including the Diploid Genetic Algorithm (DGA), using the gray matter obtained from whole-brain volume. A total of 33 experiments were conducted, incorporating various pipelines that combine the above methods. The inability of 2-D feature extraction methods to grasp the correlation between the slices can be handled by 3-D feature extraction methods, which are computationally very expensive. To handle this, the reduction of dimensions along the third axis followed by some of the above combinations has also been explored. The Genetic Algorithms and its variant Diploid Genetic Algorithm works well for selecting features extracted from various feature extraction techniques. Further, it also compares the recital of the network with various feature extraction techniques. It was observed that the combination of DWT and LBP-rotation invariant performs better than others.