What sensor accuracy is required?

While many parameters are used to determine the right sensor choice for a specific application, one of the most commonly required is accuracy. This single, seemingly straightforward term is easily understood but also frequently misunderstood, and is one way to determine whether a sensor technology or design meets the application requirements. Of course, many other factors come into play (accuracy over what measurement range, temperature range, and more), including other specifications (such as precision and repeatability), the application, market segment, industry standards, government regulations, and more.

According to the U.S. National Institute of Standards and Technology (NIST), accuracy is “the closeness of a measured value to the true value, including the concepts of bias and precision, and is judged concerning the use to be made of the data.”

Software to determine accuracy

With artificial intelligence playing a greater role in many applications, especially in healthcare, it is essential to determine the system’s accuracy required to perform properly and avoid errors and incorrect decisions. Figure 1 shows how the sensor’s initial position in the data flow impacts the final results.

One critical area for healthcare measurements is stability and the risk of falling. According to the U.S. Centers for Disease Control and Prevention (CDC), falls are the leading cause of fatal and nonfatal injuries among older adults. It also indicates that the rising number of deaths from falls among older adults can be addressed by screening for fall risk and intervening to address risk factors.

To address these concerns, researchers utilized the ZIBRIO Stability Scale. This simulation-based method provides a tool for characterizing sensor accuracy requirements in a device that employs a machine-learning algorithm to generate a postural stability (PS) score. Developed in the U.S. Space Program, the Zibrio Stability Scale provides a stability score ranging from 1 to 10. It predicts who (stability-challenged individuals) is likely to fall within the next 12 months. A score of 1 to 3 is a high fall risk, 4 to 6 indicates a moderate risk, and 7 to 10 indicates a low risk.

Postural stability scores are reported as integers ranging from 1 (worst balance) to 10 Brios (best balance), determined by an algorithm that uses 60 seconds of center-of-pressure (COP) data from 3600 samples of dynamic load cell measurements. Figure 2 shows the process and the sensor that initiates the assessment.

Based on the researchers’ simulations, a tolerable error of ±3 mm was selected for COP measurements in the Zibrio Stability Scale device.

The researchers’ approach to establishing the required application accuracy for stability assessment could provide a process for others to use or emulate for their specific measurement accuracy determination.

References

Concepts of Accuracy, Precision, and Statistical Control
Older Adult Falls Data
Zibrio Press Kit
Characterizing sensor accuracy requirements in an artificial intelligence-enabled medical device – ScienceDirect