Infrared spectroscopy allows population-level health screening

23 Jul 2024

Blood plasma testing from LMU and Max-Planck-Institute of Quantum Optics detects chronic health conditions.

Infrared spectroscopy is increasingly established as a way to carry out in vitro diagnostic testing, thanks to its ability to detect the molecular profiles of many biomarkers.

What has been missing, however, is the application of IR fingerprinting in a large-scale naturally variable population, according to a project at Ludwig-Maximilians-Universität München (LMU) and the Max Planck Institute of Quantum Optics (MPQ).

The project, based in LMU's Broadband Infrared Diagnostics (BIRD) center, aimed to assess whether combining IR fingerprinting and machine learning could produce a single-measurement multi-phenotype medical diagnostics platform. The results were published in Cell Reports Medicine.

The study builds on previous research by LMU and MPQ into using IR techniques to monitor blood composition for diagnostic purposes.

In 2021 the two institutions investigated whether IR molecular fingerprints created from blood samples via Fourier-transform infrared (FTIR) spectroscopy changed with the normal passage of time, an important consideration before proposing to use the data for health monitoring. The spectroscopic info was found to be "remarkably stable" given natural variations in most people's normal lifestyle.

The new project applied FTIR spectroscopy to 5,184 samples from 3,169 individuals within the KORA cohort, a comprehensive health research project already established in Augsburg, Germany. Machine learning methods were applied to decode the measured spectra looking for signs of dyslipidemia, hypertension, type 2 diabetes and other common conditions.

Forecasting health issues before they become apparent

"Profiling the largest population with IR fingerprinting to our knowledge, our study aimed to examine the capacity of the approach to comprehensively detect several medically relevant human health phenotypes," commented the project in its paper. "Previous clinical spectroscopy studies typically relied on rather small patient cohorts, hindering the potential of developing IR- based diagnostic models that adeptly generalize to larger populations."

Results showed that the combination of IR fingerprinting and machine learning can accurately single out healthy individuals and characterize chronic multimorbid states, according to the team. Data also showed that the approach has the capacity to forecast health problems before they become apparent.

"This new approach doesn't just pinpoint one condition at a time, it accurately identifies a range of health issues," noted LMU. "This machine learning-powered system not only identifies healthy individuals but also detects complex conditions involving multiple illnesses simultaneously. Moreover it can predict the development of metabolic syndrome years before symptoms appear, providing a window for interventions."

Although it remains to be seen whether IR fingerprinting can decode health trajectories over decade-scale time periods, the new study has validated the technique in a population-level study and points to IR molecular fingerprinting taking its place as a routine part of health screening, commented the team. This could be especially important for metabolic disorders like cholesterol abnormalities and diabetes, where timely interventions significantly improve outcomes.

"The combination of IR spectroscopy with machine learning is set to transform health diagnostics," said LMU. "With a single drop of blood and infrared light, there will be a powerful new tool to keep tabs on our health, catching problems more efficiently and potentially improving healthcare globally."