Detection of malaria parasites in dried human blood spots using mid-infrared spectroscopy and logistic regression analysis

Abstract

Background: Epidemiological surveys of malaria currently rely on microscopy, polymerase chain reaction assays (PCR) or rapid diagnostic test kits for Plasmodium infections (RDTs). This study shows that mid-infrared (MIR) spectroscopy coupled with supervised machine learning could constitute an alternative method for rapid malaria screening, directly from dried human blood spots. Methods: Filter papers containing dried blood spots (DBS) were obtained from a cross-sectional malaria survey in twelve wards in south-eastern Tanzania in 2018/19. The DBS were scanned using attenuated total reflection-Fourier transform infrared (ATR-FTIR) spectrometer to obtain high-resolution MIR spectra in the range, 4000 cm-1 to 500 cm-1. The spectra were cleaned to compensate for atmospheric water vapor and CO2 interference bands and used to train different classification algorithms to distinguish between malaria-positive and malaria-negative DBS papers based on PCR test results as reference. The analysis considered 296 individuals, including 123 PCR-confirmed malaria-positives and 173 negatives. Model training was done using 80% of the dataset, after which the best-fitting model was optimized by bootstrapping of 80/20 train/test stratified splits. The trained models were evaluated by predicting Plasmodium falciparum positivity in the 20% validation set of DBS. Results: Logistic regression was the best-performing model. Considering PCR as reference, the models attained overall accuracies of 92% for predicting P. falciparum infections (specificity = 91.7%; sensitivity = 92.8%) and 85% for predicting mixed infections of P. falciparum and P. ovale (specificity = 85%, sensitivity = 85%) in the field-collected specimen. Conclusion: These results demonstrate that mid-infrared spectroscopy coupled with supervised machine learning (MIR-ML) could be used to screen for malaria parasites in dried human blood spots. The approach could have potential for rapid and high-throughput screening of Plasmodium infections in both non-clinical settings (e.g. field surveys) and clinical settings (diagnosis to aid case management). However, full utility will require further advances in classification algorithms, field validation of this technology in other study sites and an in-depth evaluation of the biological basis of the observed test results. Training the models on larger datasets could also improve specificity and sensitivity of the technique. The MIR-ML spectroscopy system is robust, low-cost, and requires minimum maintenance.

Infrared Spectroscopy of Blood

“…To this end, two independent field trials have been undertaken in South Eastern Tanzania 124 and Thailand, 123 respectively. Mwanga et al 124 applied mid-IR spectroscopy coupled with supervised machine learning to screen malaria infection in dry blood spots obtained from 296 different volunteers. Samples were pre-analyzed with PCR and found to consist of 123 malaria positives and 173 malaria negatives.…”

A Review on Spatial Technologies for Enhancing Malaria Control: Concepts, Tools, and Challenges

“…Given the fact that more and more studies apply the deep learning algorithms for features extraction and construction [48], [49] and classification, a large amount of annotated image repositories is highly required for training to improve the classification accuracy. Mwanga et al proposed an approach that applies mid-infrared spectroscopy and logistic regression analysis to detect malaria parasites in dried human blood spots [16].…”

“…Image processing can be used to process the imageries captured using remote sensing devices [26]. In addition to that, image classification is also applied in classifying blood for humans infected by malaria disease [16].…”

Recent evolutions of machine learning applications in clinical laboratory medicine

“…Heraud et al [77] demonstrated that infrared (IR) analysis of packed RBCs using a portable spectrometer and a cloud-based SVM system could have the potential to become an accurate POC tool for the diagnosis of malaria in endemic countries. Furthermore, another study [78] also employed IR spectroscopy in combination with supervised ML to screen for malaria parasites in human dried blood spots. The authors found that Logistic Regression appeared to be the best performing model with overall accuracies of 92% for predicting Plasmodium falciparum infections and 85% in predicting mixed infections of Plasmodium falciparum and Plasmodium ovale.…”

“…While MALDI-TOF MS is commonly used for microbial identification in the routine clinical microbiology lab, vibrational spectroscopy (i.e. IR and Raman spectroscopy) is gaining considerable interest as an alternative technique in the classification of clinically relevant microorganisms due to its rapid, objective, nondestructive, and cost-effective nature [6,[76][77][78][79][80][81]. Although traditional ML techniques have been widely applied for spectral data analysis [8,[98][99][100][101][102], fewer efforts have been made in developing neural networks and deep learning algorithms [103][104][105].…”