A New Approach for Chagas Disease Screening Using Serum Infrared Spectroscopy and Machine Learning Algorithms
Matthews Martins1, Ângelo Antônio Oliveira Silva2,3, Felipe Silva Santos de Jesus2,3
1Department of Physiological Sciences, Federal University of Espírito Santo, Av. Mal. Campos, 1468-Maruípe, Vitória, Espírito Santo 29047-105, Brazil.
ACS Infectious Diseases
|August 28, 2025
Summary
This study shows that combining Fourier-transform infrared (FTIR) spectroscopy with machine learning (ML) offers a promising new method for diagnosing Chagas disease (CD). This approach could provide a cost-effective alternative to current diagnostic tests, especially in resource-limited areas.
Area of Science:
- Spectroscopy and Analytical Chemistry
- Computational Biology and Bioinformatics
- Infectious Diseases and Epidemiology
Background:
- Chagas disease (CD) impacts millions globally, with diagnosis challenging due to low parasite levels and limitations of current serological tests.
- Geographic spread of CD is increasing due to human migration.
- There is a critical need for improved, accessible diagnostic tools for Chagas disease.
Purpose of the Study:
- To evaluate the diagnostic capability of attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy combined with machine learning (ML) for Chagas disease.
- To compare the performance of ATR-FTIR/ML under dry and wet serum sample analysis conditions.
- To assess the potential of this novel approach as a cost-effective diagnostic alternative.
Main Methods:
- Serum samples from 100 individuals (49 CD-positive, 51 controls) were analyzed using ATR-FTIR spectroscopy in both dry and wet formats.
- Spectral data were processed using various machine learning algorithms (logistic regression, PLS-DA, random forest, XGBoost) for classification.
- Model performance was evaluated using accuracy, F1-score, and area under the receiver operating characteristic curve (AUC).
Main Results:
- Logistic regression achieved 93% accuracy and F1-score on dry samples, while XGBoost yielded 87% accuracy and F1-score on wet samples.
- The area under the ROC curve was notably high: 0.99 for dry and 0.92 for wet analyses.
- Permutation tests confirmed the robustness and reliability of the developed classification models.
Conclusions:
- ATR-FTIR spectroscopy coupled with ML presents a highly promising diagnostic strategy for Chagas disease.
- This method demonstrates potential as an efficient and cost-effective alternative to conventional serological assays, particularly for resource-constrained settings.
- Further validation with larger cohorts is necessary to confirm clinical applicability and specificity for widespread adoption in Chagas disease management.
