Abstract
Introduction
Infections affect 1–3% of Total Knee Arthroplasty (TKA) patients with severe ramifications to mobility. Unfortunately, reinfection rates are high (∼15%) suggesting improved diagnostics are required. A common strategy to treat TKA infection in North America is the two-stage revision procedure involving the installation of a temporary spacer in the joint while the infection is treated for 6–12 weeks before permanent revision. Subdermal temperature increases during infection by 1–4°C providing a potential indicator for when the infection has been cleared. We propose an implantable temperature sensor integrated into a tibial spacer for telemetric use. We hypothesized that suitable sensing performance for infection monitoring regarding precision and relative accuracy can be attained using a low power, compact, analog sensor with <0.1ºC resolution.
Materials & Methods
An experimental sensor was selected for our implanted application due to its extremely low (9 μA) current draw and compact chip package. Based upon dynamic range it was determined that the analog/digital converter must be a minimum of 11 bits to deliver suitable (<0.1ºC) resolution. A 12-bit ADC equipped microcontroller was selected. The MCP9808 (Microchip Technology, Chandler, AZ, USA) delivers manufacturer characterized thermal data in decimal strings through serial communication to the same microcontroller. The rated accuracy of the MCP9808 sensors in the required temperature range is max/typ +/− 0.5/0.25ºC with a precision of +/− 0.05ºC delivered at a resolution of 0.0625ºC. Within a thermally insulated chamber with a resistive heating element, the following experiment was conducted: Using empirical plant modelling tools, simulation and implementation an effective PI control scheme was implemented to create a highly precise temperature chamber. With MCP9808 as reference, the temperature in the thermal chamber was driven to 20 different temperatures between 35 and 40ºC for 10 minutes each and sampled at 5 Hz. This trial was repeated three times over three days. Transient data was discarded so as only to evaluate the steady state characteristics, wavelet denoising was applied, and a regression between the reference MCP9808 temperature response vs the experimental sensor intended for implantation was tabulated in Matlab.
Results
Compared to reference values, the experimental temperature sensor displayed relative accuracy of +/− 0.275ºC (with 95% confidence) and precision of +/−0.135ºC over a 35–40ºC range as determined over 190,212 relevant samples. Note that in practice, the precision is independent of reference, but the absolute accuracy is relative to the gold standard's accuracy.
Conclusion
Infection frequently results in permanent mobility issues in the context of total knee arthroplasty. This has led to an ongoing call for better treatments. Analysis suggests that the proposed experimental sensor offers high precision and reasonable relative accuracy in temperature sensing, substantially tighter than the expected stimulus from infection, while also offering desirable characteristics for implantation. This sensing platform will be integrated into an instrumented tibial spacer in future work.
For any figures or tables, please contact authors directly.
