A New Approach for Chagas Disease Screening Using Serum Infrared Spectroscopy and Machine Learning Algorithms

Abstract

Chagas disease (CD) affects an estimated 6–7 million people worldwide, predominantly in Latin America. However, migration has expanded its geographic reach. Diagnosing chronic CD is challenging due to low parasitemia and the limitations of existing serological assays. This study evaluates the diagnostic potential of attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy combined with machine learning (ML). A total of 100 serum samples (49 CD-positive, 51 negative controls) were analyzed using ATR-FTIR spectroscopy under two conditions: (i) dry analysis (air-dried samples) and (ii) wet analysis (direct serum analysis). Spectral data were processed using ML algorithms, including logistic regression (LR), partial least-squares discriminant analysis (PLS-DA), random forest (RF), and extreme gradient boosting (XGBoost) for sample classification. The best-performing models were LR for dry data set (accuracy and F1-score: 93%) and XGBoost for the wet data set (accuracy and F1-score: 87%). The area under the receiver operating characteristic (ROC) curve (AUC) was 0.99 and 0.92 for the dry and wet data sets, respectively. The robustness and reliability of the model were confirmed through permutation tests. These results demonstrate that ATR-FTIR spectroscopy combined with ML is a promising diagnostic tool for CD. Despite the study’s limited sample size, results suggest this approach could serve as a cost-effective alternative to conventional serological assays, particularly in resource- constrained settings. Further validation with larger data sets and diverse control groups is essential to assess its specificity and clinical applicability. If successful, this method could significantly enhance early diagnosis and improve disease managements strategies for CD.