VigilanceAI

Improving the health and safety of older adults through 'computer vision on the edge' driven monitoring services, preventing hospitalization due to UTI, Pneumonia and falls.

Comment

Inspiration

The founding team comprises of physician who specializes in older adult care, product manager who has launched AI products with gross revenue exceeding $20M / year, bootstrapped and has run multiple profitable ventures, CTO who has developed and launched computer vision models on edge devices had a common experience - keeping our older parents safe.

Driven by personal experiences and a shared vision of keeping our older parents safe, we the founders of VigilanceAI have been working to create a platform that uses advanced AI and computer vision to monitor and detect early signs of health issues including UTI, Pneumonia, preventing hospitalizations.

The founding team of VigilanceAI is not only driven by their professional expertise but also by their personal experiences with the stubborn independence of aging parents and grandparents. Witnessing the stress and challenges their own parents endured while striving to live independently in their later years has deeply influenced their shared goals. This personal connection fuels their dedication to creating a platform that leverages advanced AI and computer vision to monitor and detect early signs of health issues, ultimately aiming to provide peace of mind and improve the quality of life for seniors and their families.

What it does

VigilanceAI is a patented health monitoring platform that utilizes advanced AI and computer vision to safeguard older adults. Through cameras and edge devices in homes, it detects early signs of health issues like UTIs, pneumonia, and potential hazards for falls.

By delivering real-time alerts and insights, VigilanceAI strengthens the patient-physician relationship with precise data for personalized care. It engages patients and caregivers by offering valuable health and environmental information, promoting proactive health management.

Moreover, VigilanceAI aids clinical research by gathering real-time data on patient behavior and health, enhancing data quality for healthcare providers and researchers. This holistic approach not only ensures safety at home but also reduces healthcare costs, providing reassurance for families and caregivers.

How we built it

VigilanceAI has been built with a strong emphasis on data privacy, security, and regulatory compliance, utilizing cutting-edge technology to ensure a seamless and effective user experience. The platform employs Nvidia GPUs on the edge to process video and image data locally, ensuring that sensitive information about patients is never sent to the cloud. This design prioritizes data privacy and security, addressing concerns about unauthorized access and data breaches​

Local Processing: Video and image data are processed locally using Nvidia GPUs, which allows for real-time analysis without the need to transfer data to the cloud. This significantly reduces the risk of data breaches and ensures that patient information remains private.

Data Privacy and Security: Edge based image and video processing: Nvidia GPUs enable real-time analysis locally, minimizing the risk of data breaches and ensuring patient privacy. It utilizes Nvidia GPUs on the edge for local processing of video and image data, eliminating the need to transfer sensitive patient information to the cloud, thus mitigating risks of data breaches. No video / image data leaves the edge device and is destroyed immediately after inference.

Regulatory Compliance: Data stored and processed by VigilanceAI is handled in compliance with healthcare regulations, ensuring that the platform meets all necessary legal and ethical standards​.

HIPAA Compliance: The backend of VigilanceAI is hosted on AWS, which offers various HIPAA-compliant services. This ensures that all data storage and processing activities meet stringent regulatory standards for healthcare data​​

Patient Consent: VigilanceAI implements mechanisms for obtaining and managing patient consent for data collection and monitoring.

User-Friendly Interface: The platform offers an intuitive design for effortless interaction by older adults and caregivers, enhancing user experience and engagement.

Scalability: Built on AWS, VigilanceAI facilitates easy scaling to monitor more patients and integrate new features as required.

Infrastructure Specifications:

Nvidia GPUs on the Edge: Leveraging Nvidia GPUs enables efficient local processing for complex computer vision tasks, ensuring real-time monitoring and critical alerts to prevent health issues and emergencies.

AWS Backend: The AWS backend offers robust, scalable, and secure data storage and processing, meeting HIPAA compliance standards for data security and regulatory requirements.

Detailed architecture specifications are available upon request.

Challenges we ran into

  • Consent from participating adults to source the video data to train the model
  • Cloud credits, GPUs on the edge required to train the computer vision model from ground up.
  • Technology talent required to build and deploy AI models on the edge device.
  • Limited availability of data and manual input requirement to annotate the data.
  • Annotation at right price point.
  • Fine tuning the models to improve accuracy and precision from various camera placement locations.
  • Our computer vision model initially struggled to distinguish between similar-looking activities, such as coughing and drinking water. To address this, we fine-tuned the model to better differentiate these activities, ensuring more accurate monitoring and detection.
  • Rapidly evolving Nvidia ecosystem
  • Cloud cost for data store, real time data ingestion and analytics can spiral up.
  • Enabling infrastructure for model validation

Challenges Faced:

  • Obtaining consent from participants for sourcing video data to train the model
  • Acquiring cloud credits and GPUs for training the computer vision model from scratch
  • Securing technology talent to develop and deploy AI models on edge devices
  • Fine-tuning models to enhance accuracy and precision across various camera placements
  • Initial difficulty in distinguishing similar activities like coughing and drinking water, addressed through model refinement for improved monitoring and detection
  • Adapting to the rapidly evolving Nvidia ecosystem
  • Managing escalating cloud costs for data storage, real-time ingestion, and analytics

Accomplishments that we're proud of

  • Prototype is being piloted with 80-85% accuracy for various activity detection.
  • Supported by AARP

What we learned

  • Managing cost-effectively while delivering capabilities.
  • Advantages of fine-tuning existing models vs training new models from the ground up.
  • Compliance and regulatory requirements.

What's next for VigilanceAI

  • Complete the pilot with 20+ patients in their household by FY 2024.
  • Launch a partnership with a large hospital system in Illinois by FY 2025-2026.
  • The five-year plan is to collect data from various patient settings and approach CMS to implement this project.
  • Obtain the necessary compliance certifications by FY 2025.
Share this project:

Updates