Using mid-infrared spectroscopy and supervised machine-learning to identify vertebrate blood meals in the malaria vector, Anopheles arabiensis

Reagent-free detection of Plasmodium falciparum malaria infections in field-collected mosquitoes using mid-infrared spectroscopy and machine learning

Field-derived metrics are critical for effective control of malaria, particularly in sub-Saharan Africa where the disease kills over half a million people yearly. One key metric is entomological inoculation rate, a direct measure of transmission intensities, computed as a product of human biting rates and prevalence of Plasmodium sporozoites in mosquitoes. Unfortunately, current methods for identifying infectious mosquitoes are laborious, time-consuming, and may require expensive reagents that are not always readily available. Here, we demonstrate the first field-application of mid-infrared spectroscopy and machine learning (MIRS-ML) to swiftly and accurately detect Plasmodium falciparum sporozoites in wild-caught Anopheles funestus, a major Afro-tropical malaria vector, without requiring any laboratory reagents. We collected 7178 female An. funestus from rural Tanzanian households using CDC-light traps, then desiccated and scanned their heads and thoraces using an FT-IR spectrometer. The sporozoite infections were confirmed using enzyme-linked immunosorbent assay (ELISA) and polymerase chain reaction (PCR), to establish references for training supervised algorithms. The XGBoost model was used to detect sporozoite-infectious specimen, accurately predicting ELISA and PCR outcomes with 92% and 93% accuracies respectively. These findings suggest that MIRS-ML can rapidly detect P. falciparum in field-collected mosquitoes, with potential for enhancing surveillance in malaria-endemic regions. The technique is both fast, scanning 60-100 mosquitoes per hour, and cost-efficient, requiring no biochemical reactions and therefore no reagents. Given its previously proven capability in monitoring key entomological indicators like mosquito age, human blood index, and identities of vector species, we conclude that MIRS-ML could constitute a low-cost multi-functional toolkit for monitoring malaria risk and evaluating interventions.

Rapid assessment of the blood-feeding histories of wild-caught malaria mosquitoes using mid-infrared spectroscopy and machine learning

BACKGROUND

The degree to which Anopheles mosquitoes prefer biting humans over other vertebrate hosts, i.e. the human blood index (HBI), is a crucial parameter for assessing malaria transmission risk. However, existing techniques for identifying mosquito blood meals are demanding in terms of time and effort, involve costly reagents, and are prone to inaccuracies due to factors such as cross-reactivity with other antigens or partially digested blood meals in the mosquito gut. This study demonstrates the first field application of mid-infrared spectroscopy and machine learning (MIRS-ML), to rapidly assess the blood-feeding histories of malaria vectors, with direct comparison to PCR assays.

METHODS AND RESULTS

Female Anopheles funestus mosquitoes (N = 1854) were collected from rural Tanzania and desiccated then scanned with an attenuated total reflectance Fourier-transform Infrared (ATR-FTIR) spectrometer. Blood meals were confirmed by PCR, establishing the 'ground truth' for machine learning algorithms. Logistic regression and multi-layer perceptron classifiers were employed to identify blood meal sources, achieving accuracies of 88%-90%, respectively, as well as HBI estimates aligning well with the PCR-based standard HBI.

CONCLUSIONS

This research provides evidence of MIRS-ML effectiveness in classifying blood meals in wild Anopheles funestus, as a potential complementary surveillance tool in settings where conventional molecular techniques are impractical. The cost-effectiveness, simplicity, and scalability of MIRS-ML, along with its generalizability, outweigh minor gaps in HBI estimation. Since this approach has already been demonstrated for measuring other entomological and parasitological indicators of malaria, the validation in this study broadens its range of use cases, positioning it as an integrated system for estimating pathogen transmission risk and evaluating the impact of interventions.

Rapid classification of epidemiologically relevant age categories of the malaria vector, Anopheles funestus

BACKGROUND

Accurately determining the age and survival probabilities of adult mosquitoes is crucial for understanding parasite transmission, evaluating the effectiveness of control interventions and assessing disease risk in communities. This study was aimed at demonstrating the rapid identification of epidemiologically relevant age categories of Anopheles funestus, a major Afro-tropical malaria vector, through the innovative combination of infrared spectroscopy and machine learning, instead of the cumbersome practice of dissecting mosquito ovaries to estimate age based on parity status.

METHODS

Anopheles funestus larvae were collected in rural south-eastern Tanzania and reared in an insectary. Emerging adult females were sorted by age (1-16 days old) and preserved using silica gel. Polymerase chain reaction (PCR) confirmation was conducted using DNA extracted from mosquito legs to verify the presence of An. funestus and to eliminate undesired mosquitoes. Mid-infrared spectra were obtained by scanning the heads and thoraces of the mosquitoes using an attenuated total reflection-Fourier transform infrared (ATR-FT-IR) spectrometer. The spectra (N = 2084) were divided into two epidemiologically relevant age groups: 1-9 days (young, non-infectious) and 10-16 days (old, potentially infectious). The dimensionality of the spectra was reduced using principal component analysis, and then a set of machine learning and multi-layer perceptron (MLP) models were trained using the spectra to predict the mosquito age categories.

RESULTS

The best-performing model, XGBoost, achieved overall accuracy of 87%, with classification accuracy of 89% for young and 84% for old An. funestus. When the most important spectral features influencing the model performance were selected to train a new model, the overall accuracy increased slightly to 89%. The MLP model, utilizing the significant spectral features, achieved higher classification accuracy of 95% and 94% for the young and old An. funestus, respectively. After dimensionality reduction, the MLP achieved 93% accuracy for both age categories.

CONCLUSIONS

This study shows how machine learning can quickly classify epidemiologically relevant age groups of An. funestus based on their mid-infrared spectra. Having been previously applied to An. gambiae, An. arabiensis and An. coluzzii, this demonstration on An. funestus underscores the potential of this low-cost, reagent-free technique for widespread use on all the major Afro-tropical malaria vectors. Future research should demonstrate how such machine-derived age classifications in field-collected mosquitoes correlate with malaria in human populations.

Key considerations, target product profiles, and research gaps in the application of infrared spectroscopy and artificial intelligence for malaria surveillance and diagnosis

Studies on the applications of infrared (IR) spectroscopy and machine learning (ML) in public health have increased greatly in recent years. These technologies show enormous potential for measuring key parameters of malaria, a disease that still causes about 250 million cases and 620,000 deaths, annually. Multiple studies have demonstrated that the combination of IR spectroscopy and machine learning (ML) can yield accurate predictions of epidemiologically relevant parameters of malaria in both laboratory and field surveys. Proven applications now include determining the age, species, and blood-feeding histories of mosquito vectors as well as detecting malaria parasite infections in both humans and mosquitoes. As the World Health Organization encourages malaria-endemic countries to improve their surveillance-response strategies, it is crucial to consider whether IR and ML techniques are likely to meet the relevant feasibility and cost-effectiveness requirements-and how best they can be deployed. This paper reviews current applications of IR spectroscopy and ML approaches for investigating malaria indicators in both field surveys and laboratory settings, and identifies key research gaps relevant to these applications. Additionally, the article suggests initial target product profiles (TPPs) that should be considered when developing or testing these technologies for use in low-income settings.

Evaluation of diffuse reflectance spectroscopy for predicting age, species, and cuticular resistance of Anopheles gambiae s.l under laboratory conditions

Mid-infrared spectroscopy (MIRS) combined with machine learning analysis has shown potential for quick and efficient identification of mosquito species and age groups. However, current technology to collect spectra is destructive to the sample and does not allow targeting specific tissues of the mosquito, limiting the identification of other important biological traits such as insecticide resistance. Here, we assessed the use of a non-destructive approach of MIRS for vector surveillance, micro diffuse reflectance spectroscopy (µDRIFT) using mosquito legs to identify species, age and cuticular insecticide resistance within the Anopheles gambiae s.l. complex. These mosquitoes are the major vectors of malaria in Africa and the focus on surveillance in malaria control programs. Legs required significantly less scanning time and showed more spectral consistence compared to other mosquito tissues. Machine learning models were able to identify An. gambiae and An. coluzzii with an accuracy of 0.73, two ages groups (3 and 10 days old) with 0.77 accuracy and we obtained accuracy of 0.75 when identifying cuticular insecticide resistance. Our results highlight the potential of different mosquito tissues and µDRIFT as tools for biological trait identification on mosquitoes that transmit malaria. These results can guide new ways of identifying mosquito traits which can help the creation of innovative surveillance programs by adapting new technology into mosquito surveillance and control tools.

Artificial intelligence (AI): a new window to revamp the vector-borne disease control

Artificial intelligence (AI) facilitates scientists to devise intelligent machines that work and behave like humans to resolve difficulties and problems by utilizing minimal resources. The Healthcare sector has benefited due to this. Mosquito-transmitted diseases pose a significant health risk. Despite all advances, present strategies for curbing these diseases still depend largely on controlling the mosquito vectors. This strategy demands an army of entomology experts for thorough monitoring, determining, and finally eradicating the targeted mosquito population. Deep learning (DL) algorithms may substitute such unmanageable processes. The current review focuses on how AI, with particular emphasis on deep learning, demonstrates effectiveness in quick detection, identification, monitoring, and finally controlling the target mosquito populations with minimal resources. It accelerates the pace of operation and data exploration on ongoing evolutionary status, tendency to feed blood, and age grading of mosquitoes. The successful combination of computer and biological sciences will provide practical insight and generate a new research niche in this study area.

Using transfer learning and dimensionality reduction techniques to improve generalisability of machine-learning predictions of mosquito ages from mid-infrared spectra

BACKGROUND

Old mosquitoes are more likely to transmit malaria than young ones. Therefore, accurate prediction of mosquito population age can drastically improve the evaluation of mosquito-targeted interventions. However, standard methods for age-grading mosquitoes are laborious and costly. We have shown that Mid-infrared spectroscopy (MIRS) can be used to detect age-specific patterns in mosquito cuticles and thus can be used to train age-grading machine learning models. However, these models tend to transfer poorly across populations. Here, we investigate whether applying dimensionality reduction and transfer learning to MIRS data can improve the transferability of MIRS-based predictions for mosquito ages.

METHODS

We reared adults of the malaria vector Anopheles arabiensis in two insectaries. The heads and thoraces of female mosquitoes were scanned using an attenuated total reflection-Fourier transform infrared spectrometer, which were grouped into two different age classes. The dimensionality of the spectra data was reduced using unsupervised principal component analysis or t-distributed stochastic neighbour embedding, and then used to train deep learning and standard machine learning classifiers. Transfer learning was also evaluated to improve transferability of the models when predicting mosquito age classes from new populations.

RESULTS

Model accuracies for predicting the age of mosquitoes from the same population as the training samples reached 99% for deep learning and 92% for standard machine learning. However, these models did not generalise to a different population, achieving only 46% and 48% accuracy for deep learning and standard machine learning, respectively. Dimensionality reduction did not improve model generalizability but reduced computational time. Transfer learning by updating pre-trained models with 2% of mosquitoes from the alternate population improved performance to ~ 98% accuracy for predicting mosquito age classes in the alternative population.

CONCLUSION

Combining dimensionality reduction and transfer learning can reduce computational costs and improve the transferability of both deep learning and standard machine learning models for predicting the age of mosquitoes. Future studies should investigate the optimal quantities and diversity of training data necessary for transfer learning and the implications for broader generalisability to unseen datasets.

Highly Efficient Use of Infrared Spectroscopy (ATR-FTIR) to Identify Aphid Species

Aphids are commonly considered to be serious pests for trees, herbaceous and cultivated plants. Recognition and identification of individual species is very difficult and is based mainly on morphological features. The aims of the study were to suggest the possibility of identifying aphids through the use of Fourier-transform infrared (FTIR) spectroscopy, and to determine which absorption peaks are the most useful to separate aphid species. Using FTIR spectroscopy, based on the chemical composition of the body, we were able to distinguish 12 species of aphid. We have shown that using nine distinct peaks corresponding to the molecular vibrations from carbohydrates, lipids, amides I and II, it is possible to accurately identify aphid species with an efficiency of 98%.

Effects of sample preservation methods and duration of storage on the performance of mid-infrared spectroscopy for predicting the age of malaria vectors

BACKGROUND

Monitoring the biological attributes of mosquitoes is critical for understanding pathogen transmission and estimating the impacts of vector control interventions on the survival of vector species. Infrared spectroscopy and machine learning techniques are increasingly being tested for this purpose and have been proven to accurately predict the age, species, blood-meal sources, and pathogen infections in Anopheles and Aedes mosquitoes. However, as these techniques are still in early-stage implementation, there are no standardized procedures for handling samples prior to the infrared scanning. This study investigated the effects of different preservation methods and storage duration on the performance of mid-infrared spectroscopy for age-grading females of the malaria vector, Anopheles arabiensis.

METHODS

Laboratory-reared An. arabiensis (N = 3681) were collected at 5 and 17 days post-emergence, killed with ethanol, and then preserved using silica desiccant at 5 °C, freezing at - 20 °C, or absolute ethanol at room temperature. For each preservation method, the mosquitoes were divided into three groups, stored for 1, 4, or 8 weeks, and then scanned using a mid-infrared spectrometer. Supervised machine learning classifiers were trained with the infrared spectra, and the support vector machine (SVM) emerged as the best model for predicting the mosquito ages.

RESULTS

The model trained using silica-preserved mosquitoes achieved 95% accuracy when predicting the ages of other silica-preserved mosquitoes, but declined to 72% and 66% when age-classifying mosquitoes preserved using ethanol and freezing, respectively. Prediction accuracies of models trained on samples preserved in ethanol and freezing also reduced when these models were applied to samples preserved by other methods. Similarly, models trained on 1-week stored samples had declining accuracies of 97%, 83%, and 72% when predicting the ages of mosquitoes stored for 1, 4, or 8 weeks respectively.

CONCLUSIONS

When using mid-infrared spectroscopy and supervised machine learning to age-grade mosquitoes, the highest accuracies are achieved when the training and test samples are preserved in the same way and stored for similar durations. However, when the test and training samples were handled differently, the classification accuracies declined significantly. Protocols for infrared-based entomological studies should therefore emphasize standardized sample-handling procedures and possibly additional statistical procedures such as transfer learning for greater accuracy.

Autofluorescent Biomolecules in Diptera: From Structure to Metabolism and Behavior

Light-based phenomena in insects have long attracted researchers’ attention. Surface color distribution patterns are commonly used for taxonomical purposes, while optically-active structures from Coleoptera cuticle or Lepidoptera wings have inspired technological applications, such as biosensors and energy accumulation devices. In Diptera, besides optically-based phenomena, biomolecules able to fluoresce can act as markers of bio-metabolic, structural and behavioral features. Resilin or chitinous compounds, with their respective blue or green-to-red autofluorescence (AF), are commonly related to biomechanical and structural properties, helpful to clarify the mechanisms underlying substrate adhesion of ectoparasites’ leg appendages, or the antennal abilities in tuning sound detection. Metarhodopsin, a red fluorescing photoproduct of rhodopsin, allows to investigate visual mechanisms, whereas NAD(P)H and flavins, commonly relatable to energy metabolism, favor the investigation of sperm vitality. Lipofuscins are AF biomarkers of aging, as well as pteridines, which, similarly to kynurenines, are also exploited in metabolic investigations. Beside the knowledge available in Drosophila melanogaster, a widely used model to study also human disorder and disease mechanisms, here we review optically-based studies in other dipteran species, including mosquitoes and fruit flies, discussing future perspectives for targeted studies with various practical applications, including pest and vector control.

Dengue disease dynamics are modulated by the combined influences of precipitation and landscape: A machine learning approach

BACKGROUND

Dengue is an endemic vector-borne disease influenced by environmental factors such as landscape and climate. Previous studies separately assessed the effects of landscape and climate factors on mosquito occurrence and dengue incidence. However, both factors concurrently coexist in time and space and can interact, affecting mosquito development and dengue disease transmission. For example, eggs laid in a suitable environment can hatch after being submerged in rain water. It has been difficult for conventional statistical modeling approaches to demonstrate these combined influences due to mathematical constraints.

OBJECTIVES

To investigate the combined influences of landscape and climate factors on mosquito occurrence and dengue incidence.

METHODS

Entomological, epidemiological, and landscape data from the rainy season (July-December) were obtained from respective government agencies in Metropolitan Manila, Philippines, from 2012 to 2014. Temperature, precipitation and vegetation data were obtained through remote sensing. A random forest algorithm was used to select the landscape and climate variables. Afterward, using the identified key variables, a model-based (MOB) recursive partitioning was implemented to test the combined influences of landscape and climate factors on ovitrap index (vector mosquito occurrence) and dengue incidence.

RESULTS

The MOB recursive partitioning for ovitrap index indicated a high sensitivity of vector mosquito occurrence on environmental conditions generated by a combination of high residential density areas with low precipitation. Moreover, the MOB recursive partitioning indicated high sensitivity of dengue incidence to the effects of precipitation in areas with high proportions of residential density and commercial areas.

CONCLUSIONS

Dengue dynamics are not solely influenced by individual effects of either climate or landscape, but rather by their synergistic or combined effects. The presented findings have the potential to target vector surveillance in areas identified as suitable for mosquito occurrence under specific climatic conditions and may be relevant as part of urban planning strategies to control dengue.

Detection and Identification of Wolbachia pipientis Strains in Mosquito Eggs Using Attenuated Total Reflection Fourier Transform Infrared (ATR FT-IR) Spectroscopy

The global fight against mosquito-borne viral diseases has in recent years been bolstered by the introduction of the endosymbiotic bacteria Wolbachia to vector populations, which in host mosquitoes suppresses the transmissibility of several viruses. Researchers engaged on this front of the battle often need to know the Wolbachia infection status of individual mosquitoes, as the intervention progresses and the mosquitoes become established in the target population. Previously, we successfully applied attenuated total reflection Fourier transform infrared spectroscopy to the detection of Wolbachia in adult Aedes aegypti mosquitoes; here we apply the same principles to Aedes eggs, with sensitivity and selectivity > 0.95. Further, we successfully distinguish between infections in eggs of the wMel and wMelPop strains of Wolbachia pipientis, with a classification error of 3%. The disruption of host lipid profile by Wolbachia is found to be a key driver in spectral differences between these sample classes.

Podoconiosis: key priorities for research and implementation

Podoconiosis is a non-infectious tropical lymphoedema causing swelling of the lower legs. Podoconiosis is associated with stigma, depression and reduced productivity, resulting in significant socio-economic impacts for affected individuals, families and communities. It is caused by barefoot exposure to soils and affects disadvantaged populations. Evidence from the past 5 y suggests that podoconiosis is amenable to public health interventions, e.g. footwear and hygiene-based morbidity management, which reduce acute clinical episodes. Although much has been learned in recent years, advances in care for these patients and worldwide control requires further reliable and relevant research. To develop a comprehensive global control strategy, the following key research priorities are important: better understanding of the global burden of podoconiosis through extended worldwide mapping, development of new point-of-care diagnostic methods and approaches to define the presence of the environmental characteristics that contribute to the development of the condition, improving treatment through an increased understanding of the pathogenesis of dermal changes over time, improved understanding of optimal ways of providing patient care at the national level, including research to optimize behavioural change strategies, determine the optimum package of care and integrate approaches to deliver robust surveillance, monitoring and evaluation of control programmes.

Predicting Infrared Spectra with Message Passing Neural Networks

Infrared (IR) spectroscopy remains an important tool for chemical characterization and identification. Chemprop-IR has been developed as a software package for the prediction of IR spectra through the use of machine learning. This work serves the dual purpose of providing a trained general-purpose model for the prediction of IR spectra with ease and providing the Chemprop-IR software framework for the training of new models. In Chemprop-IR, molecules are encoded using a directed message passing neural network, allowing for molecule latent representations to be learned and optimized for the task of spectral predictions. Model training incorporates spectra metrics and normalization techniques that offer better performance with spectral predictions than standard practice in regression models. The model makes use of pretraining using quantum chemistry calculations and ensembling of multiple submodels to improve generalizability and performance. The spectral predictions that result are of high quality, showing capability to capture the extreme diversity of spectral forms over chemical space and represent complex peak structures.

Detection of Plasmodium falciparum in laboratory-reared and naturally infected wild mosquitoes using near-infrared spectroscopy

There is an urgent need for high throughput, affordable methods of detecting pathogens inside insect vectors to facilitate surveillance. Near-infrared spectroscopy (NIRS) has shown promise to detect arbovirus and malaria in the laboratory but has not been evaluated in field conditions. Here we investigate the ability of NIRS to identify Plasmodium falciparum in Anopheles coluzzii mosquitoes. NIRS models trained on laboratory-reared mosquitoes infected with wild malaria parasites can detect the parasite in comparable mosquitoes with moderate accuracy though fails to detect oocysts or sporozoites in naturally infected field caught mosquitoes. Models trained on field mosquitoes were unable to predict the infection status of other field mosquitoes. Restricting analyses to mosquitoes of uninfectious and highly-infectious status did improve predictions suggesting sensitivity and specificity may be better in mosquitoes with higher numbers of parasites. Detection of infection appears restricted to homogenous groups of mosquitoes diminishing NIRS utility for detecting malaria within mosquitoes.

The application of spectroscopy techniques for diagnosis of malaria parasites and arboviruses and surveillance of mosquito vectors: A systematic review and critical appraisal of evidence

CONCLUSIONS/SIGNIFICANCE

The potential of RS as a surveillance tool for malaria and arbovirus vectors and MIRS for the diagnosis and surveillance of arboviruses is yet to be assessed. NIRS capacity as a surveillance tool for malaria and arbovirus vectors should be validated under field conditions, and its potential as a diagnostic tool for malaria and arboviruses needs to be evaluated. It is recommended that all 3 techniques evaluated simultaneously using multiple machine learning techniques in multiple epidemiological settings to determine the most accurate technique for each application. Prior to their field application, a standardised protocol for spectra collection and data analysis should be developed. This will harmonise their application in multiple field settings allowing easy and faster integration into existing disease control platforms. Ultimately, development of rapid and cost-effective point-of-care diagnostic tools for malaria and arboviruses based on spectroscopy techniques may help combat current and future outbreaks of these infectious diseases.

A Literature Review of Host Feeding Patterns of Invasive Aedes Mosquitoes in Europe

Aedes invasive mosquitoes (AIMs) play a key role as vectors of several pathogens of public health relevance. Four species have been established in Europe, including Aedes aegypti, Aedesalbopictus, Aedes japonicus and Aedes koreicus. In addition, Aedes atropalpus has been repeatedly recorded although it has not yet been established. In spite of their importance in the transmission of endemic (e.g., heartworms) and imported pathogens (e.g., dengue virus), basic information of parameters affecting their vectorial capacity is poorly investigated. The aim of this study is to review the blood feeding patterns of these invasive mosquito species in Europe, summarizing available information from their native and introduced distribution ranges. The feeding patterns of mosquitoes constitute a key parameter affecting the contact rates between infected and susceptible hosts, thus playing a central role in the epidemiology of mosquito-borne pathogens. Our results highlight that these mosquito species feed on the blood of different vertebrate groups from ectotherms to birds and mammals. However, humans represent the most important source of blood for these species, accounting for 36% and 93% of hosts identified for Ae. japonicus and Ae. aegypti, respectively. In spite of that, limited information has been obtained for some particular species, such as Ae. koreicus, or it is restricted to a few particular areas. Given the high vector competence of the four AIM species for the transmission of different emerging arboviruses such as dengue, Chikungunya, Zika or Yellow fever viruses and their high feeding rates on humans, these AIM species may have an important impact on the vectorial capacity for such pathogens on urban and periurban areas. Finally, we propose directions for future research lines based on identified knowledge gaps.

Assessment of vector-host-pathogen relationships using data mining and machine learning

Infectious diseases, including vector-borne diseases transmitted by arthropods, are a leading cause of morbidity and mortality worldwide. In the era of big data, addressing broad-scale, fundamental questions regarding the complex dynamics of these diseases will increasingly require the integration of diverse datasets to produce new biological knowledge. This review provides a current snapshot of the systematic assessment of the relationships between microbial pathogens, arthropod vectors and mammalian hosts using data mining and machine learning. We employ PRISMA to identify 32 key papers relevant to this topic. Our analysis shows an increasing use of data mining and machine learning tasks and techniques, including prediction, classification, clustering, association rules mining, and deep learning, over the last decade. However, it also reveals a number of critical challenges in applying these to the study of vector-host-pathogen interactions at various systems biology levels. Here, relevant studies, current limitations and future directions are discussed. Furthermore, the quality of data in relevant papers was assessed using the FAIR (Findable, Accessible, Interoperable, Reusable) compliance criteria to evaluate and encourage reproducibility and shareability of research outcomes. Although shortcomings in their application remain, data mining and machine learning have significant potential to break new ground in understanding fundamental aspects of vector-host-pathogen relationships and their application in this field should be encouraged. In particular, while predictive modeling, feature engineering and supervised machine learning are already being used in the field, other data mining and machine learning methods such as deep learning and association rules analysis lag behind and should be implemented in combination with established methods to accelerate hypothesis and knowledge generation in the domain.

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

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 investigated whether 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 12 wards in southeastern 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 vapour 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 Plasmodium 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 human DBS. The approach could have potential for rapid and high-throughput screening of Plasmodium in both non-clinical settings (e.g., field surveys) and clinical settings (diagnosis to aid case management). However, before the approach can be used, we need additional field validation in other study sites with different parasite populations, and in-depth evaluation of the biological basis of the MIR signals. Improving the classification algorithms, and model training on larger datasets could also improve specificity and sensitivity. The MIR-ML spectroscopy system is physically robust, low-cost, and requires minimum maintenance.