DYNASPECTRUM. Spectrometry signal analysis for exposome characterisation through AI techniques

Dynaspectrum is a project funded under the 2022 Multi-Area Internal Projects call in which the Digital Health and Biomedical Devices Area and the Energy, Industrial Processes and Environment Area collaborate.

Context

The Volatil Organics Compounds characterisation has multiple applications such as detection of uncompleted combustion processes, leakages of gases in industrial contexts, identification of diseases or control of ecosystems’ health.

Nowadays, there are air quality stations in Euskadi that incorporate sensors for measuring some specific Volatil Organic Compounds (or VOCs). For example, Benzece in Muskiz station. However, only few stations incorporate these kinds of sensors, and even when measuring VOCs, only some of them are covered.

 Technological Challenges

Mass spectrometry facilitates the analysis of the exposome because it provides values of VOCs, distinguishing the patterns of different chemical compounds. The question is about:

• The usefulness and quality of a miniaturize mass spectrometry device, called SPOC
• About the possible integration of SPOC data with the data of air quality stations in Euskadi into a common application. 
• And about the scalability of such un application and its usefulness in different contexts

Methodology

The methodological approach includes:

• An empirical study that gets real world data able to characterize the exposome
• ETL + MLOPS methodologies for the development of deep learning analysis
• Agile software development able to integrate different sources of data, predictive models, and visualization

Technical Results

DYNASPECTRUM project has shown good results for the characterisation of Volatile Organic Compounds (VOCs):

• When recognizing Acetone, we used a well controlled environment (Figure8) and stablish a baseline with an accuracy and precision of 98%
• When also found good results when the SPOC was used to detect different conditions at indoor and outdoor environments with an accuracy above 88% 
• Finally, when trying to identify different chemical mixtures at industrial, rural and urban contexts (Figure 12) we visit Mazarredo Muskiz and Arraiz and got an accuracy above 83%. These results were improved when integrating the information available at Open Data Euskadi.

Scientific Results

• Registration of SOLARIS library for the analysis of signals by using deep learning strategies. It includes specific tools for data pre-processing and analysis for AI based classification, regression and interpretability. 
• IEEE International Symposium on Circuits and Systems (IEEE ISCAS 2025) in London on the 25-28th of May 2025. Paper submitted to the “Environmental Impacts on Healthcare and Wellbeing and Active Interventions” special issue. 
• Participation in a HE preparation proposal: ENACT - Environmental Effect on Health Care and Wellbeing and Active Interventions. Funded. Already working at the GA and CA preparation.

Applications

The identification of Volatile Organic Compounds has several applications for the exposome characterisation, and includes all those related to Health management, environmental monitoring, and wellbeing and daily routines.

We acknowledge the support of TECHNION for providing SPOC technology generously 

Watch the YouTube video