The graph on the left shows a schematic of the data fed into the neural network, which are related to each other and to other hidden variables to obtain body temperature as an output. On the right is a graph showing the results.
Other studies have demonstrated the usefulness of skin temperature as a parameter, for example, being able to predict future infections in hospitalised patients [6].
2.- Solution proposed by Intelligent Data: Customised calibration.
In the first stage of the pilot phase of the project, it is proposed to perform a personalised measurement calibration for each patient. That is, when a patient is given the bracelet, the skin temperature is automatically taken by the bracelet and, simultaneously, the nursing staff measures the body temperature. The difference between them is a first calibration of the personalised and unique increment for each patient:
Δ(𝑡 = 0) = 𝑇𝑐(0) - 𝑇𝑝(0)
The increment is variable over time and several calibrations can be performed during the patient's stay in hospital.
In the following stages of the pilot phase, the calibration can be adjusted by including the following variables in its calculation which, as this document has shown, affect the time variation of the customised increment:
- Ambient temperature.
- Heart rate.
The values of these variables are collected by the wristband at the same time as the skin temperature, so there would be no time lag between them.