Inspiration
Our project was inspired by the work of Dr. Peii Chen and Dr. Joan Toglia, who developed the 3S-v2 Test for assessing spatial neglect in individuals with brain injuries (Chen & Toglia, 2022). Their research systematically examines both egocentric (viewer-centered) and allocentric (stimulus-centered) spatial neglect, providing valuable insights into cognitive assessment methods. This research highlighted the importance of precise and functional assessments in cognitive healthcare, inspiring us to develop a digital, secure approach for continuous cognitive monitoring. Our solution aims to make cognitive health assessments both accessible and private, supporting practitioners and researchers in monitoring cognitive changes over time.
What it Does
The 3s Spatial Awareness Assessment securely encrypts and stores cognitive test results, enabling private and continuous cognitive assessments without compromising user privacy. Every test result remains encrypted end-to-end, ensuring that data cannot be accessed by unauthorized parties. The encrypted results are used to refine and improve the AI model, allowing it to adapt to new data trends and improve assessment accuracy over time. This setup ensures a dynamic, self-improving model that supports users’ long-term cognitive health monitoring in a secure and scalable way.
How We Built It
The backend was built with Python, incorporating PyTorch for model training and TenSEAL for secure computations. The model employs CKKS encryption, which allows exact arithmetic on encrypted data, ideal for processing sensitive cognitive data.
Challenges We Ran Into
One of the primary challenges was implementing CKKS in a way that supports efficient computation on encrypted data without compromising accuracy. CKKS is computationally intensive, and handling it within machine learning workflows poses challenges around performance and scalability. Another challenge was developing a model that could continuously learn from encrypted data, as updating the model while ensuring data remains secure demanded careful architecture planning and optimization.
Accomplishments That We're Proud Of
This project demonstrates that privacy doesn’t have to come at the expense of innovation. We also achieved a significant milestone in integrating encrypted data within an AI model that improves over time, providing a novel approach to privacy-preserving machine learning.
What We Learned
This project deepened our understanding of encryption techniques. Moreover, building an adaptive AI model with encrypted data required us to rethink traditional machine learning workflows and innovate to handle the unique constraints of privacy-preserving data processing.
What's Next for 3s Spatial Awareness Assessment
With future investment and research, we plan to expand the capabilities of the 3s Spatial Awareness Assessment by refining its AI model for even greater accuracy and exploring more efficient encryption schemes to improve processing speeds. Additional features, such as personalized cognitive health insights and integration with other health data sources, could enhance the tool's value in long-term cognitive health monitoring.
Log in or sign up for Devpost to join the conversation.