How artificial intelligence is shaping neuropsychology: A focus on cognitive assessment of neurodegenerative disorders

TL;DR

This paper reviews how AI and machine learning are changing cognitive assessments for brain disorders like Alzheimer's and Parkinson's.

Contribution

The paper evaluates ML algorithms' accuracy and proposes ways to improve their clinical use in neuropsychology.

Findings

ML algorithms like SVMs and CNNs are being used to classify neurodegenerative disorders using biomarkers and tests.

Current ML methods struggle with overfitting and distinguishing between similar pathologies.

The paper suggests new research directions to enhance ML accuracy and clinical relevance in cognitive assessments.

Abstract

Artificial intelligence (AI) and machine learning (ML) algorithms are revolutionising the world, and they have the potential to revolutionise neuropsychology as well. A particularly fruitful field for this revolution is the cognitive assessment of neurodegenerative disorders, such as Alzheimer's disease, Parkinson's disease, Mild Cognitive Impairment and Primary Progressive Aphasia. This narrative review explores the impact of ML and AI in classifying these patients by using biomarkers or neuropsychological tests, using vast amounts of data and providing previously unattainable insights. Additionally, the article will evaluate the accuracies of several ML algorithms, such as support vector machines, random forest or convolutional neural networks. The article will also discuss the challenges related to ML like the risk of overfitting and the need for ML algorithms to execute a differential analysis among several pathologies—a capability that current research has yet to achieve fully. Furthermore, it proposes new directions to improve the clinical utility and accuracy of ML classification algorithms in neuropsychology, underlining the possibility for theoretical advancements based on the results of these classifications.