Distance-based novelty detection model for identifying individuals at risk of developing Alzheimer's disease

• DOI: 10.3389/fnagi.2024.1285905
• Publication Date: 2024-04-15T13:53:15Z
• Authors: Hongqin Yang; Jiangbing Mao; Qinyong Ye; Magda Bucholc; Shuo Liu; Wenzhao Gao; Jie Pan; Jiawei Xin; Xuemei Ding
• Keywords: Graphical user interface; Novelty Detection

Introduction Novelty detection (ND, also known as one-class classification) is a machine learning technique used to identify patterns that are typical of the majority class and can discriminate deviations novelties. In context Alzheimer's disease (AD), ND could be employed detect abnormal or atypical behavior may indicate early signs cognitive decline presence disease. To date, few research studies have risk developing AD mild impairment (MCI) from healthy controls (HC). Methods this work, two distinct cohorts with highly heterogeneous data, derived Australian Imaging Biomarkers Lifestyle (AIBL) Flagship Study Ageing project Fujian Medical University Union Hospital (FMUUH) China, were employed. An innovative framework built-in easily interpretable models constructed solely on HC data was introduced along proposing strategy distance boundary (DtB) MCI AD. Subsequently, web-based graphical user interface (GUI) incorporates proposed developed for non-technical stakeholders. Results Our experimental results best overall performance detecting individuals in AIBL FMUUH datasets obtained by using Mixture Gaussian-based algorithm applied single modality, an AUC 0.8757 0.9443, sensitivity 96.79% 89.09%, specificity 89.63% 90.92%, respectively. Discussion The GUI offers interactive platform aid stakeholders making diagnoses AD, enabling streamlined decision-making processes. More importantly, DtB visually quantitatively at