Boosting Medical Ultrasound Technology

Witekio began by conducting a thorough review of the startup's existing codebase and system architecture. The goal was to assess the foundation and ensure the system could handle the required AI workloads. This involved:

• Auditing both the MCU and MPU environments.
• Identifying any security or performance bottlenecks.
• Providing a comprehensive roadmap to make the system scalable and production-ready.

With a solid understanding of the system's limitations, Witekio set out to update the operating environments:

• MCU Environment: The team integrated real-time operating systems (RTOS) to ensure low-latency processing for anomaly detection.
• MPU Environment: Witekio transitioned the existing Linux system to a Yocto-based build for improved security, customization, and performance. This provided a flexible base for running AI models on the i.MX 8 processor.

Witekio's expertise in AI was critical to the project's success. The solution involved two key components:

• AI on the MCU: For hardware anomaly detection, Witekio integrated a simple yet powerful AI model on the microcontroller. This model monitored various sensors (current, temperature, etc.) to detect abnormalities, helping predict potential hardware failures and ensuring the device’s longevity.
• AI on the MPU: Using the industry-standard framework TensorFlow Lite, Witekio enabled real-time video processing on the i.MX 8 processor. This allowed the ultrasound device to run AI models locally, assisting in medical diagnostics by processing images directly on the device. The AI model was designed to be updated and iterated upon, ensuring continuous improvement and flexibility in future diagnostics.