[2J1-GS-10a] AI application

9:45 AM - 10:00 AM JST(12:45 AM - 1:00 AM UTC)

[2J1-GS-10a-04] Generation of Texture Images for Dementia ScreeningValidation of AI-Generated Stimuli Based on Onomatopoeic Evaluation

〇Yuji Nozaki¹, Ryota Abe¹, Chihiro Kamohara², Madoka Nakajima², Maki Sakamoto¹ (1. The University of Electro-Communications, 2. Juntendo University)

Keywords:

Dementia,Onomatopoeia,Image Generation

This study explores the feasibility of using AI-generated texture images as visual stimuli for dementia screening based on onomatopoeic responses. While early detection of cognitive decline is essential, existing screening methods often place psychological and cognitive burdens on examinees. Texture-based screening using onomatopoeia has been proposed as an alternative; however, collecting sufficient real-world texture images is costly and labor-intensive. To address this issue, we focused on “fluffy” textures, which have shown potential in previous studies, and generated texture images using an image generation AI. Subjective evaluation experiments were conducted with young healthy participants, who rated the perceived texture of the images on a five-point Likert scale. Additionally, image feature analysis was performed to assess bias within the generated image set. The results indicate that the AI-generated images achieved acceptable perceptual validity, suggesting their potential use as experimental stimuli in texture-based cognitive assessments. This study highlights the utility of generative AI for efficiently expanding texture image datasets and supports its application in the development of scalable dementia screening methods.