This project focuses on improving the accessibility and inclusivity of hospital environments for visually impaired individuals. Centered on Spain’s Primary Care Centres (CAPs), the initiative addresses critical navigation challenges faced by this community. By leveraging AI-driven tools, including topological graph analysis and semantic segmentation, the project provides architects with actionable insights to design healthcare spaces that are safer, more efficient, and aligned with accessibility guidelines.
Research and analysis
The research highlighted a steady global increase in vision impairment due to conditions such as cataracts and diabetic retinopathy. Despite this, healthcare spaces, especially CAPs, fail to meet accessibility standards, creating unnecessary challenges for visually impaired users.
Navigation Challenges in Healthcare: Among all building typologies, healthcare facilities are notably complex and challenging to navigate, with Spain’s Primary Care Centres (CAPs) posing particular barriers.
Architectural Neglect: A significant gap exists in the focus on designing spaces that genuinely cater to visually impaired individuals.
Our Solution
To address this gap, we’ve developed an AI-powered assessment tool tailored for architects to enhance accessibility in healthcare environments, focusing on Spain’s CAPs.
Core Processes of the Solution
Guideline Compliance
Our tool ensures alignment with established accessibility standards, offering actionable insights for improvement
Topological Graph Analysis
Using advanced algorithms like those in TopologicPy, spatial relationships within architectural layouts are analyzed:
- Adjacency Detection: Identifying connected spaces.
- Connectivity Analysis: Mapping interactions within the building.
- Centrality Measures: Highlighting critical navigation points such as Degree, Betweenness, and Eigenvector Centrality.
- Pathfinding with Dijkstra’s Algorithm: Creating efficient, safe paths for visually impaired users.
Semantic Segmentation
Architectural layouts are annotated using tools like Roboflow, identifying elements such as doors, walls, and furniture. Combined with accessibility guidelines and NLP models like GPT, the platform generates JSON-formatted design suggestions for enhanced inclusivity.
Why AI and Graph Theory?
Our deterministic graph-based approach eliminates the need for machine learning in this static context. Algorithms ensure precise, reliable outputs, achieving over 99% pathfinding accuracy and 85–95% segmentation precision, depending on data quality.
Our second process emphasizes the integration of visual data into our platform to generate actionable design insights. Architectural layouts are converted into annotated images using tools like Roboflow, with key elements such as doors, walls, floors, and furniture meticulously labeled. These annotated visuals are combined with accessibility guidelines, which are structured and analyzed using advanced Natural Language Processing (NLP) models like GPT. This fusion enables the generation of JSON-formatted recommendations, offering architects precise, practical solutions to enhance accessibility.
For instance, the system may propose optimized tactile marker placements or safer, more intuitive navigation routes, directly improving the inclusivity of architectural designs.
Our approach leverages well-established algorithms and tools to ensure precision and reliability.
- Topological Graph Analysis: Graph-based algorithms are deterministic, meaning the outputs are repeatable and consistent.
- Tools like NetworkX ensure robust implementations for spatial analysis, achieving an accuracy of over 99% for pathfinding tasks.
- Image Segmentation with Roboflow: For well-labeled datasets, segmentation accuracy typically ranges between 85–95%, depending on task complexity and image quality.
- NLP Analysis with GPT: GPT models are highly effective at interpreting structured guidelines and generating actionable insights. For well-structured input, accuracy can reach 90–95%
In summary, our project demonstrates how AI can bridge the gap between architectural design and accessibility for visually impaired users. By combining topological graph analysis, image segmentation, and natural language processing (NLP), we provide architects with an innovative platform to design more inclusive spaces.
This solution addresses both technical and practical challenges, ensuring that hospital environments, such as CAPs, are safer, more efficient, and easier to navigate.
References:
https://www.ada.gov/law-and-regs/design-standards
https://www.ada.gov/law-and-regs/design-standards
https://www.healthfacilityguidelines.com
https://topologic.app/topologicpy_doc/topologic_pdoc/index.html
