A literature review was conducted to analyse the correlation between VOCs in breath and various diseases. Spectral simulations were performed to find suitable absorption lines that allow laser-based detection while avoiding cross-sensitivities to other gases present in the breath such as water and carbon dioxide. The prototype’s architecture was developed and the specifications derived. Mitigating the risk that the highly innovative technologies impose on the successful integration to a functional prototype, more established technologies were selected to enable the use of the breath analyser in the clinical study. In parallel, the innovative technologies are further developed in an explorative prototype, enabling the potential to achieve a break-through and contributing to provide a more compact, powerful breath analyser.
The clinical settings for the validation of the VOCORDER device have been defined. The clinical study comprises a baseline and a validation phase to achieve significant results in terms of exploring the correlation of biomarkers to the selected diseases using mass spectroscopy as reference method and evaluating the impact of this new breath analysis technology.
WP3 focuses on developing essential components for the final prototype. High-heat load DFB-QCL lasers at targeted wavelengths have been designed, with a five-wavelength fiber-packaged DFB array in progress. The DFB array and multipass cell have been simulated and designed, and fabrication is underway, with identified risks addressed. A rotating mirror will combine laser beams for the multipass cell, and initial system calculations have been completed. For the self-mixing scheme, a theoretical model for quantum cascade lasers under optical feedback has guided an experimental setup with promising results. Electronic subsystems are being integrated on schedule, with software development proceeding as planned.
WP4 focuses on developing artificial intelligence (AI) models for disease detection using VOCORDER’s breath and EMR data, including AI algorithms for identifying disease indicators in spectral biomarkers. WP4 involves data pre-processing, feature selection, and model training with fuzzy AI to boost detection accuracy. Significant progress has been made in scientific technological objectives of WP4. EMR data integration was achieved using HL7/FHIR standards, securing anonymized, compliant data transfer. An in-depth analysis of circadian rhythms was carried out, establishing that AI models are capable of detecting time-dependent variations in volatile organic compound data from breath samples. Progress in Cystic Fibrosis diagnosis yielded 88.07% predictive accuracy, with optimal feature sets achieving 100%. SHAP-based AI models enhance transparency, marking a step toward AI-powered breath analysis.
W5 aims at delivering the integrated VOCORDER system and at validating it in laboratory environment. The integration is split in two tasks covering the integration of optoelectronic components with the electronics and software and the integration of cloud-based data analytics using AI algorithms. An integration plan is developed deriving the timeframe for the stepwise integration of the components. For the integration, the expertise and resources of the partners are considered to create synergies and realize an efficient way to integrate the individual components into functional modules. Eventually, the integrated prototype will be validated in laboratory settings where a dedicated validation setup is established enabling a reproducible and highly accurate evaluation of the VOCORDER device performance using precise gas mixtures.
