Autonomous Ultrasound AI for Extreme Environments
Biomedical Engineering
Andrew Booth, Lakshya Dharwal, Genevieve Oakes, Camryn Sheen, Isabella Soriano, Hunter Weber
Abstract
Access to diagnostic imaging is a major limitation in spaceflight and remote medical environments, where trained clinicians and cloud connectivity are unavailable. To address this critical risk, Celestia was developed as a tablet-based diagnostic imaging platform capable of autonomous, real-time AI-assisted interpretation without external computational resources. Our design occupies a unique market position by resolving the core engineering conflict between power efficiency and diagnostic performance, enabling high-quality imaging in highly resource-constrained environments. This capability is essential for ensuring medical self-sufficiency during long-duration missions, where rapid and accurate diagnostic insight can significantly reduce clinical uncertainty.
The system meets stringent NASA mission constraints by targeting Max Power Draw ≤ 0.8W and achieving an exceptional ≤ 3 second diagnostic latency. This ultra-efficiency is enabled by integrating a lightweight convolutional neural network (CNN) optimized via TensorFlow Lite on an Android-based tablet. This architecture achieves the required clinical efficacy, targeting ≥ 90% Sensitivity and Specificity for key diagnostics, including free fluid detection. The guided user interface provides visual and auditory feedback, allowing non-expert users, such as astronauts, to capture clinically useful images autonomously. These interface features reduce
operator dependency and help ensure consistent imaging quality even under stressful or unfamiliar conditions.
Celestia leverages commercial off-the-shelf hardware and edge-optimized AI inference to achieve high performance with low cost and complexity. Preliminary analysis estimates a prototype manufacturing cost of $1,200-$1,500 USD. The design is developed in alignment with key medical device standards, including IEC 60601 (safety), ISO 14971 (risk management), and IEC 62366 (usability), aligning with a Class II FDA 510(k) regulatory pathway. By providing reliable, autonomous diagnostic power, Celestia enhances readiness for both space exploration and critical terrestrial telemedicine applications.
Video
Research poster
Faculty mentor
Partner
