LG-FIS: A Hierarchical Local-Global Fuzzy Inference System for Multimodal Alzheimer’s Disease Diagnosis

The timely and precise diagnosis of the Alzheimer’s disease problem is highly important because clinical manifestations are heterogeneous, and the data is unclear. As a means of multi-stage diagnosis of Alzheimer, the present paper suggests a hierarchical Local-Global Fuzzy Inference System (LG-FIS) that is a combination of clinical, cognitive and genetic modalities. All the modalities are fuzzified separately to produce local risk indices, which then are integrated using a global fuzzy decision layer. The experiments have shown higher accuracy in classification, excellent Early MCI detection, and healthy resistance to noisy inputs. The fuzzy framework is based on rules, and therefore, the fuzzy framework is highly interpretable to support the assessment of the Alzheimer's disease with clarity and clinical significance.