Multi-Modal Feature Fusion with Fuzziness based Semi-Supervised Learning for Alzheimer’s Disease Diagnosis

Alzheimer’s disease is a prominent kind of dementia that is caused by a slow decline of cognitive abilities, including memory loss, impaired speech and diminished reactivity. The condition primarily affects brain regions responsible for cognition, memory and language and typically becomes apparent after the age of 60, with a greater risk among older individuals. To meet the requirement for early identification, researchers have explored various machine-learning techniques. To enhance the accuracy of diagnosing Alzheimer’s disease, researchers have employed supervised and unsupervised learning approaches derived from machine learning. While supervised learning relies on labelled data, obtaining sufficient labelled samples can be time-consuming and requires specialized assistance. The focus is on multiclass classification, harnessing MRI and PET data. SGLDM and SIFT feature extraction algorithms are utilized to enable the fusion of features from diverse perspectives, resulting in a robust classification model. This research introduces a novel strategy for AD diagnosis, combining multimodal feature fusion with a fuzziness-based-semi-supervised learning method. The method leverages both labeled and mislabeled data to enhance the model’s reliability. The proposed method demonstrated notable enhancements in the classifier’s performance compared to existing approaches on the dataset.