Blood gas measurements are a vital tool in monitoring the health of prematurely born children. About three quarters of all patients in neonatal intensive care units are monitored in this way to prevent morbidity and mortality, particularly from hyperoxia and hypoxia. However, current methods and technology suffer from major problems in both handling and reliability, which in the end jeopardises the quality of care for among the most vulnerable patients in modern healthcare. These problems include lack of continuous monitoring, heating of the skin that causes burns, slow response times that cannot detect emergencies, and the fact that current apparatus only works for patients with sufficiently thin skin (neonates) while patients with thicker skin (children and adults) too would benefit greatly from the technique. The need for development in the field of transcutaneous blood gas monitoring is well acknowledged by stakeholders in both industry and healthcare.

This project will investigate how novel microplasma emission spectrometry can be used to solve all of the above problems and, hence, create a breakthrough in the quality of current neonatal care. The solution builds on miniaturised sensor technology that originally was developed for planetary exploration, more precisely to look for signs of past or percent life on Mars. This sensor has several unique traits that are ideal for so-called transcutaneous, i.e., through the skin, blood gas measurements, which is a technique that analyses the minute amount of gas that diffuses from the blood through the skin. The sensor’s miniature embodiment makes it able to study extremely small total sample amounts and, hence, can perform the analysis much faster than the competition. Moreover, the measurement scheme, where the sample is analysed by studying the light it emits after being ionised, offers great sensitivity.

However, the output data is quite complex and achieving sufficient reliability requires detailed analysis. This project aims to solve this issue by incorporating machine learning algorithms in the software. Machine learning is ideal for handling big and complex data sets and we believe that it will not only enable us to greatly improve the precision and accuracy but also extend the scope of the analysis to more complex questions.

In a second study, we will construct a stand-alone prototype and incorporating the new software. This prototype will be benchmarked against the latest instruments from the market-leading companies to confirm its performance with respect to the state of the art. These studies will, in the end, be published in open source scientific journals and presented at the ATTRACT Final Event.