Semantic verbal fluency assessment using computational analysis in the Czech language.

Area of Science:

• Neuropsychology
• Computational linguistics
• Neurology

Background:

• Semantic verbal fluency (SVF) tasks are crucial in neuropsychological assessment.
• Declines in SVF performance are indicative of mild cognitive impairment (MCI) and Alzheimer's disease (AD).
• Traditional qualitative SVF analysis has limitations and yields mixed results.

Purpose of the Study:

• To evaluate the diagnostic potential of a computational method for MCI assessment in the Czech Republic.
• To compare the novel computational approach with traditional qualitative analysis of SVF tasks.
• To analyze Number of Switches (NOS) and Mean Cluster Size (MCS) in animal and vegetable categories.

Main Methods:

• Analysis of NOS and MCS in semantic verbal fluency tasks (animal and vegetable categories).
• Comparison between traditional qualitative analysis and a novel computational approach.
• Assessment of diagnostic value using Area Under the Curve (AUC) for NOS and MCS.

Main Results:

• In the animal category, NOS showed fair diagnostic value computationally, while MCS had poor value by both methods.
• In the vegetable category, NOS demonstrated fair diagnostic value for both traditional (AUC=0.761) and computational (AUC=0.708) methods.
• No significant differences in diagnostic value were found between computational and traditional approaches for most indexes.

Conclusions:

• Computational analysis of SVF tasks is a promising tool for MCI assessment in the Czech context.
• The computational method shows comparable diagnostic potential to traditional qualitative analysis.
• Further refinement is needed to address limitations like group imbalances and multi-word phrase handling.