Machine Learning in Cognitive Neuroscience: A Promising Approach for Early Detection of Alzheimer's Disease

Background: Alzheimer's disease (AD) is the most common cause of dementia, affecting a significant number of individuals globally. AD is classified as a subtype of dementia, a comprehensive term that encompasses a decline in cognitive abilities and behavioural changes to the extent that it significantly disrupts everyday activities (Orru et al., 2009; Coin et al., 2009). In this context, early detection of AD is crucial for timely intervention and treatment, enabling more effective symptoms management and potentially slowing the disease's progression (Dubois et al., 2016). In recent year, machine learning (ML) techniques have become evident as a promising avenue to evaluate brain imaging data and detect characteristics linked to different neurological or psychiatric conditions (i.e., Dwyer, Falkai, & Koutsouleris, 2018; Ferrucci et al., 2022) and it has been successfully applied in various fields, including forensic sciences (i.e., Pace et al., 2019; Mazza et al., 2020), psychological research (i.e., Orrù et al., 2021; 2023) and liver transplantation (i.e., Ferrarese et al., 2021), amongst others. Aim: the objective of this study was to provide a brief overview of the current state of research on the application of ML in cognitive neuroscience for the early detection of AD. Methods/Data Sources: we conducted a literature search of studies published in peer-reviewed journals. We included some of the studies that applied ML techniques to neuroimaging data, cognitive assessments, or multimodal data for the purpose of early detection of AD. Results: Moradi et al., (2014) introduced an innovative magnetic resonance imaging (MRI)-derived technique for forecasting the progression MCI to AD up to three years prior diagnosis. The authors have created a “novel MRI biomarker”. This biomarker when combined with age and cognitive assessments, resulted in a comprehensive biomarker, referred to as the aggregate biomarker, which demonstrated a 10-fold cross-validated AUC of 0.9020 in accurately distinguishing between progressive and stable MCI patients. In another study, Khedher et al., (2014) have developed and introduced a computer-aided diagnosis system for early AD detection. The system was able to achieve promising results in distinguishing AD and MCI patients from healthy controls (sensitivity = 85.11%, specificity = 91.27% and accuracy 88.49%). Interestingly, the study conducted by Uddin et al., (2023) applied a ML model which included different algorithms to accurately predict the occurrence of AD. The model was trained using the open access OASIS dataset. The results revealed that the Voting Classifier achieved the highest level of validation accuracy (96%). Salvatore et al., (2015) demonstrated that ML algorithms applied to structural brain MRI data have yielded encouraging results in discriminating among AD patients, MCI patients who will convert to AD (MCIc), MCI patients who will not convert to AD (MCInc), and healthy controls (AD vs healthy controls, 76%; MCIc vs healthy controls, 72%; MCIc vs MCInc, 66%). The study conducted by Nanni et al., (2020) attempted to assess the effectiveness of ensemble transfer-learning methods in predicting early diagnosis and prognosis of AD. The results demonstrated that the ensemble transfer-learning approach effectively discriminated between AD and healthy controls, MCIc and healthy controls, and MCIc and MCInc. The study by Venugopalan et al., (2021) used deep learning (DL) models combined with multiple data. The study revealed that deep models had superior performance compared to shallow models. Conclusions: based on the literature presented, ML algorithms and deep learning models have shown promising results in analysing complex patterns in neuroimaging data, cognitive assessments, and other biomarkers to identify individuals at risk of developing AD.