Case Study: Medical Device

                                                   With Edge based Machine Learning solution

Product

  • Embedded and encapsulated medical device for use in patient monitoring.

  • Based on NXP NHS31xx (NXP reference design shown).

  • Battery powered.

  • Designed to wake periodically to collect and store patient data.

NHS31xx Block Diagram, Ref. NXP.

Challenges

  • Contracted with a large established medical device company - this was however their first electronic / software based product.

  • Firmware Challenges

    • The device and firmware needed to meet their extreme low power and size constraints.

      • Device placed in-patient.

      • Data collected and logged over weeks.

    • Inconsistent Data Sets

      • Issues with finding patterns in the data for decision making.

  • Product Test Challenges

    • Comprehensive product testing was needed.

  • FDA Regulatory Requirements

    • Processes and tools needed for FDA compliance.

Firmware Solution

  • Low Power Implementation

    • Extensive profiling and analysis of Deep Sleep and Active Mode to minimize current draw.

  • Machine Learning (ML)

    • Implemented an Edge based Machine Learning solution to address uniqueness in data sets.

    • Model Training was done off chip.

    • Inference implemented on device.

Product Test Solutions

  • In house Functional Test, Engineering Test and Manufacturing Test environments were defined and implemented.

  • Python used to implement 24/7 automated Functional Testing.

  • Included Web interfaces for data logging (MySQL / MariaDB) and test reporting.

FDA Regulatory Solutions

  • Full implementation of required FDA procedures and documentation.

  • This gave the client a clear path to FDA Pre-Market Approval (PMA).

© 2020 Golden Gate Research Inc.