Distinguishing age-related cognitive decline from dementias: A study based on machine learning algorithms

Area of Science:

• Neuroscience
• Artificial Intelligence
• Gerontology

Background:

• Age-related cognitive decline (ARCD) presents challenges in differentiation from various dementia types.
• Accurate diagnosis is crucial for appropriate patient management and treatment strategies.

Purpose of the Study:

• To investigate the efficacy of machine learning in distinguishing ARCD from probable Alzheimer's disease (AD), vascular dementia (VD), and amnestic mild cognitive impairment (MCI).
• To identify key neurocognitive tests that contribute to accurate classification.

Main Methods:

• 106 subjects were categorized into ARCD, AD, VD, and MCI groups.
• A comprehensive battery of neurocognitive tests was administered, including memory, executive function, and language assessments.
• Machine learning algorithms (multilayer perceptron, support vector machine, M5-model trees) were employed for classification.

Main Results:

• ARCD was 100% separable from AD, 95.4% from MCI, and 86.3% from VD.
• Öktem verbal memory processes (Ö-VMPT) recognition, Ö-VMPT total learning, semantic fluency, proverbs, Stroop interference, and Boston Naming Test (BNT) were identified as highly discriminative tests.

Conclusions:

• Machine learning models demonstrate significant potential for accurately differentiating ARCD from various dementia syndromes.
• Neurocognitive test data, when analyzed with machine learning, can provide a robust basis for distinguishing between normal aging and pathological cognitive decline.