Exploring Visual Complaints through a test battery in Acquired Brain Injury Patients: A Detailed Analysis of the DiaNAH Dataset

Gonçalo Hora de Carvalho · 2025

Read Exploring Visual Complaints through a test battery in Acquired Brain Injury Patients: A Detailed Analysis of the DiaNAH Dataset by Gonçalo Hora de Carvalho (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-18540.txt.

Introduction

Gonçalo Hora de Carvalho studies the visual complaints of people living with acquired brain injury. This 2025 paper works with the DiaNAH dataset, which covers 948 patients who filled in a questionnaire on eight kinds of visual symptoms. These include blurred vision and trouble with contrast, and many records have gaps.

After dropping incomplete cases, 767 patients remained for analysis, and machine learning filled the missing values in a careful way. Patients were grouped by their complaint patterns, which can combine in tens of thousands of ways. The study found little direct link between what patients report and what standard vision tests measure. The author calls for larger studies to untangle how felt symptoms and test scores relate after brain injury.

Catalog record

CREATORGonçalo Hora de Carvalho
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-18540

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Exploring Visual Complaints through a test battery in Acquired Brain Injury Patients: A Detailed Analysis of the DiaNAH Dataset Gonçalo Hora de Carvalho Abstract This study investigated visual impairment complaints in a sample of 948 Acquired Brain Injury (ABI) patients using the DiaNAH dataset, emphasizing advanced machine learning techniques for managing missing data. Patients completed a CVS questionnaire capturing eight types of visual symptoms, including blurred vision and altered contrast perception. Due to incomplete data, 181 patients were excluded, resulting in an analytical subset of 767 individuals. To address the challenge of missing data, an automated machine learning (AutoML) approach was employed for data imputation, preserving the distributional characteristics of the original dataset. Patients were grouped according to singular and combined complaint clusters derived from the 40,320 potential combinations identified through the CVS questionnaire. A linear correlation analysis revealed minimal to no direct relationship between patient-reported visual complaints and standard visual perceptual function tests. This study represents an initial systematic attempt to understand the complex relationship between subjective visual complaints and objective visual perceptual assessments in ABI patients. Given the limitations of sample size and variability, further studies with larger populations are recommended to robustly explore these complaint clusters and their implications for visual perception following brain injury. Journal: Elsevier Introduction The human brain, in its incredible complexity, orchestrates the perception, understanding, and interaction with the world around us. The capacity of this organ to process vast quantities of information is facilitated, in no small part, by our sense of vision. However, injury to the brain can have a significant impact on visual perception and related functions and previous studies have demonstrated that visual symptoms are prevalent among Acquired Brain Injuries (ABIs) patients and can persist long after the initial injury (Wang et al., 2021). This study seeks to further explore the utility of the DiaNAH dataset - a visual test battery on ABI patients. One of the challenges of studies involving human participants is the construction of consistent and comprehensive datasets. Researchers also often face the issue of missing data, which can occur for many reasons, from patients not completing certain sections of a survey to technical glitches in data collection apparatuses (Smith & Jones, "Challenges in Medical Data Collection"). The field of ABIs has seen considerable growth in the last decade with the advent of datasets such as the Traumatic Brain Injury Model Systems (TBIMS) National Database (Miller et al., "The TBIMS National Database: Advancements and Future Directions"). However, the DiaNAH dataset, with its focus on visual complaints post-ABI, occupies a unique niche. Given its size, of the main challenges in utilizing the DiaNAH dataset is dealing with missing data. Missing data can occur for various reasons, including incomplete survey responses and technical issues during data collection. This issue can significantly impact the reliability of the analysis if not appropriately addressed. Traditional methods for handling missing data, such as list-wise deletion or mean imputation, can lead to biased results and reduced statistical power (Rubin, 2004). Advanced machine learning techniques for data imputation may provide a more reliable alternative, preserving the dataset’s integrity and allowing for more accurate analysis (Van Buuren & Groothuis-Oudshoorn, 2011). These machine learning algorithms can predict and fill in missing values based on the available data, thereby maintaining the dataset’s overall structure and variability. This approach is particularly useful in medical research, where incomplete data is a common challenge (Sterne et al., 2009). The analysis of the DiaNAH dataset involved then examining the relationships between visual complaints and various visual perceptual function tests. This study aimed to identify potential clusters of complaints and explore their correlations with the test results. Our motivation to do this was to understand these relationships so as to deduce underlying mechanisms of visual impairments in ABI patients and inform the development of targeted interventions. Previous research has highlighted the complexity of visual complaints in ABI patients, suggesting that these complaints may not always correlate directly with measurable visual deficits (Geldmacher et al., 2020). For instance, a study by Miller et al. (2019) found that while many ABI patients report significant visual complaints, these are not always reflected in standard visual perceptual tests. This discrepancy underscores the need for a more nuanced approach to analyzing visual complaints, considering both subjective reports and objective test results. The DiaNAH dataset, together with a CVS questionnaire and extensive battery of visual tests, provides a unique opportunity to delve deeper into these issues. By employing data analysis techniques and machine learning algorithms, this study seeks to uncover patterns and correlations that might not be apparent through traditional methods. The findings from exploring the DiaNAH dataset could have significant implications for the assessment and management of visual impairments in ABI patients, ultimately improving their quality of life. This study represents an initial effort to systematically explore and preprocess the DiaNAH dataset in order to enable further exploration of the complex interplay between visual complaints and visual perceptual function in ABI patients, paving the way for future research in this area. Methodology The DiaNAH dataset comprises patients who suffer from ABI, each of which completed a computer vision syndrome questionnaire (CVS) questionnaire and a visual impairment test battery. The CVS questionnaire is a tool utilized to gather patient data on phenotypes, encompassing both symptoms and causes. The total number of ABI patients within the DiaNAH dataset is 948. However, due to a range of exclusion criteria that resulted in missing values (further elaborated for each test under Appendix - Neuropsychological Tests and Evaluations ), some patients were precluded from the analysis. In total, 181 patients were omitted, resulting in an analytical dataset comprised of 767 patients (an in-depth description follows in Subsection Missing Values ). In the context of this study, complaints were defined as patient responses to specific questions about their general visual health. The subsequent table illustrates these complaints: Complaint question Question CVS 12 Do you have blurred vision? CVS 13 Do you have trouble seeing letters in a text against the background? CVS 14 Are you blinded by bright light more than before the brain injury? CVS 15 Do you have the impression that everything seems darker (do you experience a dark haze) than before the brain injury? CVS 16 Do you need more light when you see than before the brain injury? CVS 18 If you go out into bright sunlight, does it take longer than before the brain injury to get used to the bright light? CVS 19 When you go back inside from the bright sunlight, does it take longer than before the brain injury to get used to it? CVS 20 Do you experience colors differently than before the brain injury? Table 1: Summary of Complaint Questions Used in the Survey. Complaint question Description Label CVS 12 Blurred vision 1000 0000 CVS 13 Contrast perception 0100 0000 CVS 14 Light nuisance 0010 0000 CVS 15 Dark haze 0001 0000 CVS 16 Increased need of light 0000 1000 CVS 18 Disrupted light adaptation 0000 0100 CVS 19 Disrupted dark adaptation 0000 0010 CVS 20 Color differences 0000 0001 Table 2: Descriptions and Labels of Complaint Questions Correlation In terms of features, the DiaNAH dataset contains a total of 389 variables. When

Read the complete text