Determining the performance of a temperature sensor embedded into a mouthguard

Intra-oral temperature monitoring is clinically important as it relates to core temperature and can indicate infections, medication reactions, and disease-specific symptoms. In dentistry, it helps diagnose periodontal disease and tooth erosion/decay and can be used to monitor compliance with oral appliances. Accurate temperature readings are essential. Various technologies (thermocouples, RTDs, semiconductor ICs) are used for intra-oral measurement, each with trade-offs. Accuracy can be affected by circuit design, sensor technology, placement, temperature range, and encapsulation. Therefore, performance of the encapsulated sensor should be assessed across a relevant temperature range, and the time to reach equilibrium (steady-state) should also be considered. The study addresses a gap by presenting a pre-clinical simulation evaluating both error and time components for mouthguard-embedded temperature sensors, based on ISO standards, complemented by a preliminary in vivo case study.

The paper outlines that thermocouples are commonly used for intra-oral temperature sensing due to wide range, low cost, and fast response; RTDs and semiconductor ICs are also used but less commonly. Prior studies typically tested reproducibility in a specific setting, pre-encapsulation accuracy, or compared readings in humans against a reference thermometer, but often did not evaluate encapsulated sensors across a clinically relevant temperature range or consider time to steady-state. Some dentistry studies used water bath tests (e.g., Farella et al. at 25, 35, 45 °C) primarily for other components (e.g., pH) rather than validating the temperature sensor. Prior work indicates water baths may not mimic intra-oral conditions, particularly for time dynamics, and wear-time monitoring sensors are often validated in thermostatic baths at a single setpoint, which may not represent oral heterogeneity. This study positions itself as the first to assess oral-based temperature sensors for long-term use across a relevant range with both error and time-to-equilibrium analyses.

Instrumentation: A flexible EVA mouthguard was fabricated via impression and casting, then pressure-molded with a 1.20 mm EVA layer. A custom data acquisition system (ARM Cortex-M4 microprocessor STM32L476JGY up to 80 MHz, BlueNRG-MS Bluetooth 4.2, Cypress S25FS512S flash memory) and a digital temperature sensor (MAX30208, ±0.1 °C from +30 to +50 °C) were secured on the first EVA layer with adhesive. The sensor was placed at the buccal side of the upper first molar to minimize airflow-induced variability. A second EVA layer encapsulated the board; edges were smoothed for comfort. Devices were tethered (wired) to the MAX30208 Evaluation System for robust data capture; Bluetooth was not used. Wires exited anteriorly; a pH-sensitive strip monitored potential water ingress. In vitro testing: Based on BS EN ISO 80601-2-56:2017 (section 201.101.2). Tests were conducted at room temperature (~21 °C). An instrumented mouthguard and an RS PRO RS1710 PT1000 wired digital thermometer were submerged in a circulating water bath controlled by a heater (Anova Nano) with ±0.1 °C accuracy. The bath was set to 34.0, 38.5, and 43.0 °C at least 30 min prior; temperature maintained ±0.1 °C. Readings from both devices were logged every 10 s for 30 min. Sensors were placed generally near the heating module; location/orientation were not fixed. Four instrumented mouthguards were tested across the three setpoints. Preliminary in vivo testing: One volunteer wore the instrumented mouthguard for 30 min. Reference: 3M Tempa DOT single-use clinical thermometers (±0.1 °C). Buccal temperature (same sensor location) was measured before insertion and after removal; sublingual temperature was measured every 5 min during wear. Mouthguard logged every 10 s. Two mouthguards were tested with the volunteer. Data analysis: Errors were computed as the difference between mouthguard readings and reference sensors (RS1710 and water bath sensor in vitro; single-use thermometers in vivo). Reported metrics: mean absolute error (MAE) and root mean squared error (RMSE). Steady-state was defined as 60 s of continuous data with variation <0.02 °C (long-term stability specification of MAX30208), corroborated by visual plateaus. For simplicity, steady-state analyses focused on data from 20–30 min. Spurious readings were filtered via localized mean and standard deviation methods. Analyses used Matlab R2019b.

In vitro: Median time to temperature equilibrium for the mouthguard was 380 s (range 130–690 s). By setpoint: 34 °C median 360 s (130–690 s); 38.5 °C median 425 s (150–510 s); 43 °C median 455 s (170–630 s). Steady-state errors (1200–1800 s): vs RS1710 thermometer, MAE (mean ± SD) across setpoints: 34 °C 0.13 ± 0.22 °C; 38.5 °C 0.18 ± 0.24 °C; 43 °C 0.24 ± 0.16 °C; total 0.21 ± 0.21 °C. RMSE: 34 °C 0.12 ± 0.25 °C; 38.5 °C 0.18 ± 0.27 °C; 43 °C 0.24 ± 0.18 °C; total 0.21 ± 0.28 °C. Versus water bath sensor, MAE: 34 °C 0.20 ± 0.20 °C; 38.5 °C 0.20 ± 0.24 °C; 43 °C 0.20 ± 0.23 °C; total 0.20 ± 0.22 °C. RMSE: 34 °C 0.20 ± 0.23 °C; 38.5 °C 0.19 ± 0.27 °C; 43 °C 0.20 ± 0.27 °C; total 0.20 ± 0.23 °C. Overall, a mean absolute error of about 0.2 °C was achieved after a maximum of 690 s across conditions. In vivo: Median time to steady-state was 1030 s (range 950–1110 s). Compared to single-use thermometers placed at the same buccal location, steady-state MAE was 0.23 ± 0.16 °C and RMSE 0.24 ± 0.21 °C. Compared to sublingual temperature (body temperature), steady-state MAE was 0.72 ± 0.11 °C and RMSE 0.72 ± 0.15 °C. Case study summary statement: absolute error about 0.2 °C after 1110 s.

Embedded temperature sensors in EVA mouthguards produced consistent steady-state accuracy (~±0.2 °C) across clinically relevant temperatures, close to clinical requirements. The encapsulation material did not appear to degrade steady-state accuracy, aligning with prior work on embedding effects. However, the time to equilibrium differed substantially between water bath and in vivo settings, with in vivo taking longer, likely due to airflow and heterogeneous temperature distribution in the oral cavity. Sensor placement and encapsulation optimization may mitigate intra-oral gradients, but designers must balance response speed with robustness to transient oral events (e.g., hot/cold beverages) that can cause large but brief temperature excursions. The study highlights that while water bath calibration across multiple setpoints is essential and informative for steady-state error, baths are a limited analog for intra-oral conditions when time dynamics matter (e.g., wear-time logging). Thus, protocols relying solely on water bath tests, especially at a single setpoint, may not capture intra-oral temporal heterogeneity.

Mouthguards instrumented with temperature sensors can meet clinical steady-state accuracy requirements across a range of temperatures. Nonetheless, time to temperature equilibrium is substantially longer in vivo compared to water bath tests, indicating water baths are not a complete representation for applications sensitive to time dynamics. A comprehensive assessment process, including both multi-setpoint bench testing and in vivo evaluation, is recommended for intra-oral temperature sensors.

Small sample size (four mouthguards in vitro; one volunteer in vivo) limits generalizability. Wired design required additional waterproofing and limited testing with more devices/subjects. Ambient room temperature before immersion was not controlled and may have influenced time to steady-state. Accuracy of reference sensors (water bath and RS1710) is of similar magnitude to the mouthguard sensor, constraining finer error analysis. Sensor placement/orientation in the water bath was not fixed, potentially introducing minor variability.