Spatial Analysis of Image Descriptions for Detection of Cognitive Disorders

Abstract

Neurodegenerative diseases pose a significant challenge to modern medicine, not onlydue to their progressive and irreversible nature but also because they often remain undiagnoseduntil symptoms become severe. Early detection is essential to maximize the effectivenessof available treatments and to slow the progression of cognitive decline. However, currentdiagnostic approaches are often slow, costly, and inaccessible, typically requiring in-personclinical assessments or expensive procedures like MRI.Research by Bartoš et. al. (2024) has proposed an alternative approach that leveragescognitively demanding tasks - such as describing complex, detailed images in combinationwith other short-term memory tests as a means of early detection.This paper explores the potential of automating the analysis of such image descriptiontasks using machine learning techniques, with the goal of developing a fast, low-cost screeningtool that can be used at home, potentially with the assistance of a family member, therebyreducing the need for in-person clinical visits. Specifically, the focus is on one specific subcomponentof this approach: analyzing the sequence of objects as mentioned during imagedescription.