A Comparative Evaluation of Artificial Intelligence and Physician Based Approaches in the Early Diagnosis of Cognitive Aging in Patients with Subjective Cognitive Decline and Risk of Alzheimer’s Disease

Background: As the society ages, the number of patients with early cognitive impairment that can progress to Alzheimer’s disease also increases. Early diagnosis and risk as-sessment allows effectively initiate the necessary lifestyle changes and monitoring. The use of artificial intelligence (AI), when analyzing medical histories, enables more pro-ductive evaluation of large datasets and identify patterns that may go unnoticed in clinical practice. This kind of approach can improve early screening, reduce physicians’ workload and develop bigger support for personalized treatment. The aim of the study: To compare the performance of machine learning (ML) algorithm with a physician (neurologist) in assessing patient’s subjective cognitive decline and Alz-heimer’s disease risk in early stages. Research methods: The research was designed as a retrospective, comparative cohort study that used two data sources. Firstly, the National Alzheimer’s Coordination Center (NACC) longitudinal dataset to train the ML model. Secondly, medical records gathered from Pauls Stradins Clinical University Hospital dating from 2020 till May 2025 to evaluate the al-gorithm’s precision. Results: The research included 154 patients, predominantly women (68.8%), with a mean age of 80.3 years. Class distribution consisted of dementia (n=139); mild cognitive im-pairment (MCI) (n=13); subjective cognitive decline (SCD) (n=2). Dementia was identified the best – 128/139 (accuracy – 92.1%) with errors tending towards MCI. MCI was correct in 9/13 cases (accuracy – 69.2%) All SCD cases were classified as dementia. Overall model’s accuracy was 91.6% (141/154). Conclusions: ML algorithm can match to neurologist made diagnoses with high precision but is struggles to separate adjacent early-stage diagnoses. At this moment, ML models are great decision supporters, but no yet alone diagnosticians. Nevertheless, this technology has high potential to being integrated in the future to aid triage and early screening, especially when advanced diagnostics are limited.