Comparison of the Accuracy of Four Malaria Diagnostic Methods in a High Transmission Setting in Coastal Cameroon

Exploring the Impact of Batch Size on Deep Learning Artificial Intelligence Models for Malaria Detection

Introduction Malaria is a major public health concern, especially in developing countries. Malaria often presents with recurrent fever, malaise, and other nonspecific symptoms mistaken for influenza. Light microscopy of peripheral blood smears is considered the gold standard diagnostic test for malaria. Delays in malaria diagnosis can increase morbidity and mortality. Microscopy can be time-consuming and limited by skilled labor, infrastructure, and interobserver variability. Artificial intelligence (AI)-based tools for diagnostic screening can automate blood smear analysis without relying on a trained technician. Convolutional neural networks (CNN), deep learning neural networks that can identify visual patterns, are being explored for use in abnormality detection in medical images. A parameter that can be optimized in CNN models is the batch size or the number of images used during model training at once in one forward and backward pass. The choice of batch size in developing CNN-based malaria screening tools can affect model accuracy, training speed, and, ultimately, clinical usability. This study explores the impact of batch size on CNN model accuracy for malaria detection from thin blood smear images. Methods We used the publicly available "NIH-NLM-ThinBloodSmearsPf" dataset from the United States National Library of Medicine, consisting of blood smear images for Plasmodium falciparum. The collection consists of 13,779 "parasitized" and 13,779 "uninfected" single-cell images. We created four datasets containing all images, each with unique randomized subsets of images for model testing. Using Python, four identical 10-layer CNN models were developed and trained with varying batch sizes for 10 epochs against all datasets, resulting in 16 sets of outputs. Model prediction accuracy, training time, and F1-score, an accuracy metric used to quantify model performance, were collected. Results All models produced F1-scores of 94%-96%, with 10 of 16 instances producing F1-scores of 95%. After averaging all four dataset outputs by batch size, we observed that, as batch size increased from 16 to 128, the average combined false positives plus false negatives increased by 15.4% (130-150), and the average model F1-score accuracy decreased by 1% (95.3%-94.3%). The average training time also decreased by 28.11% (1,556-1,119 seconds). Conclusion In each dataset, we observe an approximately 1% decrease in F1-score as the batch size was increased. Clinically, a 1% deviation at the population level can create a relatively significant impact on outcomes. Results from this study suggest that smaller batch sizes could improve accuracy in models with similar layer complexity and datasets, potentially resulting in better clinical outcomes. Reduced memory requirement for training also means that model training can be achieved with more economical hardware. Our findings suggest that smaller batch sizes could be evaluated for improvements in accuracy to help develop an AI model that could screen thin blood smears for malaria.

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.

Non-falciparum species and submicroscopic infections in three epidemiological malaria facets in Cameroon

BACKGROUND

There are growing reports on the prevalence of non-falciparum species and submicroscopic infections in sub-Saharan African countries but little information is available from Cameroon.

METHODS

A hospital-based cross-sectional study was carried out in four towns (Douala, Maroua, Mayo-Oulo, and Pette) from three malaria epidemiological strata (Forest, Sahelian, and Soudanian) of Cameroon. Malaria parasites were detected by Giemsa light microscopy and polymerase chain reaction (PCR) assay. Non-falciparum isolates were characterized and their 18S gene sequences were BLASTed for confirmatory diagnosis.

RESULTS

PCR assay detected malaria parasites in 82.4% (98/119) patients, among them 12.2% (12/98) were asymptomatic cases. Three Plasmodium species viz. P. falciparum, P. ovale curtisi and P. vivax, and two co-infection types (P. falciparum + P. vivax and P. falciparum + P. ovale curtisi) were found. The remaining infections were mono-infections with either P. falciparum or P. ovale curtisi. All non-falciparum infections were symptomatic and microscopic. The overall proportion of submicroscopic infections was 11.8% (14/119). Most asymptomatic and submicroscopic infection cases were self-medicated with antimalarial drugs and/or medicinal plants. On analysis, P. ovale curtisi sequences were found to be phylogenetically closer to sequences from India while P. vivax isolates appeared closer to those from Nigeria, India, and Cameroon. No G6PD-d case was found among non-falciparum infections.

CONCLUSIONS

This study confirms our previous work on circulation of P. vivax and P. ovale curtisi and the absence of P. knowlesi in Cameroon. More studies are needed to address non-falciparum malaria along with submicroscopic infections for effective malaria management and control in Cameroon.

Current status of 4-aminoquinoline resistance markers 18 years after cessation of chloroquine use for the treatment of uncomplicated falciparum malaria in the littoral coastline region of Cameroon

The onset and rapid spread of chloroquine resistance and the introduction of amodiaquine for the treatment of uncomplicated malaria in Cameroon have influenced the proportion of Plasmodium falciparum sensitive and resistant alleles related to 4-aminoquinoline drugs. This study was undertaken to determine the prevalence of resistance markers to antimalarial 4-aminoquinolines in Douala in the Littoral Region, and Buea in the South West Region in June 2020. Dry blood spots were prepared from malaria microscopy positive cases and used for parasite DNA extraction by chelex-100 method. Plasmodium species identification was carried out by PCR amplification/agarose gel electrophoresis of 18srRNA. The Pfcrt and Pfmdr1 genes were amplified by PCR followed by restriction digestion. The prevalence of single nucleotide polymorphisms (SNPs) was compared between study sites and with previous studies carried out between 2003-2005 and 2009-2011 using the Chi square test. The results showed that Plasmodium falciparum was the dominant species occurring as mono-infections (84.6%). The wild type K76 allele of the Pfcrt gene was found in 74.9% of isolates while the wild N86, Y184 and D1246 alleles of the Pfmdr1 gene were found respectively in 87.2%, 89.6% and 100% of field isolates. The results showed a significant reduction in the mutant alleles compared to results obtained in 2003-2005 and 2009-2013. The KNYD haplotype was observed to be the most prevalent. The results indicated that there is a gradual erosion of the mutant Pfcrt and Pfmdr1 genotype and a gradual return to the sensitive P. falciparum genotype in Cameroon.

Comparison of Rapid Diagnostic Test, Microscopy, and Polymerase Chain Reaction for the Detection of Plasmodium falciparum Malaria in a Low-Transmission Area, Jazan Region, Southwestern Saudi Arabia

This cross-sectional study aimed to assess the performances of a rapid diagnostic test (RDT)—the AllTest Malaria p.f./p.v., microscopy, and nested polymerase chain reaction (PCR) for diagnosing Plasmodium falciparum malaria in 400 febrile patients from a low-transmission region (Jazan) in southwestern Saudi Arabia. Diagnostic performance of all three methods was compared using microscopy and nested PCR as reference methods. Overall, 42 (10.5%), 48 (12.0%), and 57 (14.3%) samples were found positive by microscopy, RDT, and PCR, respectively. With PCR as reference method, the RDT showed higher sensitivity (79% vs. 71.9%), similar specificity (99.1% vs. 99.7%), and better NLR (0.20 vs. 0.27) and area under the curve (89.0% vs. 85.8%) than microscopy. The sensitivity of RDT and microscopy decreased as age increased, and false negatives were associated with low parasite density. In addition, the sensitivity of RDT and microscopy was higher in non-Saudi than in Saudi participants. Against microscopy, both RDT and PCR showed high sensitivity (83.3% vs. 97.6%), specificity (96.4% vs. 95.5%), and NPVs (98.0% vs. 99.7%), but reduced PPVs (72.9% vs. 71.9%), respectively. The results showed that the performance of the AllTest Malaria p.f./p.v RDT was better than that of microscopy in diagnosing P. falciparum malaria among febrile patients in the Jazan region when nested PCR was used as the reference. However, further studies are required to assess malaria diagnostic methods among asymptomatic individuals in the region.

Diagnosis of malaria in pregnancy: accuracy of CareStart™ malaria Pf/PAN against light microscopy among symptomatic pregnant women at the Central Hospital in Yaoundé, Cameroon

Background

The need to start treatment early for pregnant women who present with clinical features of malaria usually conflicts with the need to confirm diagnosis by microscopy (MP) before treatment, due to delays in obtaining results. Parasite sequestration in the placenta is also a problem. Rapid diagnostic tests (RDT), which detect soluble antigens, are a valuable alternative. The objective of this study was to evaluate pretreatment parasite prevalence by microscopy and by RDT and to assess the accuracy of RDT with MP as reference.

Methods

A prospective cross-sectional study was carried out at the obstetrical unit of the Central Hospital in Yaoundé, during the period January-August 2015. Consenting patients with symptoms of suspected malaria in pregnancy were recruited and a blood sample taken for MP and RDT before treatment was started. The estimates of diagnostic performance (with 95% confidence interval) were calculated in OpenEpi online software using the Wilson’s score. The agreement, as reflected by the Cohen’s kappa, was calculated and interpreted using known intervals.

Results

The results showed that, out of the 104 patients recruited, 69.2% (95%CI: 59.1–77.5) were MP positive while 77.94% (95%CI: 63.1–80.9) were RDT positive. The sensitivity of the malaria RDT was 91.67% (95%CI: 83.69–96.77) while the specificity was 53.13% (95%CI: 31.39–65.57). The diagnostic accuracy of the RDT with MP as reference was 79.81% (95%CI: 70.0–86.1). All cases were due to Plasmodium falciparum. A Cohen’s kappa of 0.45 (95%CI: 0.26–0.64) was obtained, consistent with a moderate agreement between the tests.

Conclusions

The diagnostic accuracy of the CareStart™ malaria Pf/PAN compared to microscopy was high, but not as desirable, with a false negative RDT at very high parasitaemia. In tertiary facilities, RDTs appear to provide a better diagnostic solution compared to microscopy. However, future studies with larger sample sizes should make this observation more generalizable; as missing a case could have serious consequences on pregnancy outcome.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12936-022-04109-6.

Ultrasensitive qPCR-Based Detection of Plasmodium falciparum in Pregnant Women Using Dried Blood or Whole Blood Pellet Samples Processed through Different DNA Extraction Methods

Highly sensitive molecular techniques for the detection of low-level Plasmodium falciparum parasitemia are highly useful for various clinical and epidemiological studies. However, differences in how blood samples are preserved, the quantity of blood stored, as well as genomic DNA extraction methods used may compromise the potential usefulness of these methodologies. This study compared diagnostic sensitivity based on microscopy and malaria rapid diagnostic tests (mRDTs), with quantitative polymerase chain reaction (qPCR) P. falciparum positivity of dried blood spots (DBS) or whole blood pellets (WBP) from pregnant women using different DNA extraction protocols (Chelex-saponin or a commercial kit). Samples from 129 pregnant women were analyzed, of which 13 were P. falciparum positive by mRDT and 5 by microscopy. By using extraction kit on WBP and on DBS, qPCR positivity was 27 (20.9%) and 16 (12.4%), respectively, whereas Chelex extraction on DBS only resulted in 4 (3.1%) P. falciparum positive samples. Thus, extraction using commercial kits greatly improve the likelihood of detecting P. falciparum infections.

Comparison of the performance of the CareStart Malaria Pf/Pan Combo test and field microscopy in the diagnosis of Plasmodium vivax malaria in North Sumatera, Indonesia

Background

In areas where malaria is endemic and where trained microscopists are not available, rapid diagnostic tests (RDTs) are needed not only to allow prompt treatment without delay but also to prevent overdiagnosis and overtreatment based on clinical judgements that may lead to drug resistance. This study aimed to compare the performances of the CareStart Pf/Pan Combo test to field microscopy, which is considered to be the gold standard for malaria diagnosis.

Methods

Any person with a fever or a history of fever within 48 h who came to the health centre was recruited for the study and tested both by the CareStart Pf/Pan test and by field microscopy. Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were analysed with both methods.

Results

Two-hundred study participants were enrolled: 96 (48%) were found to be positive through microscopy, while 100 (50%) participants were found to be positive through RDT. The RDT produced four false-positive results. High sensitivity and specificity were observed for the CareStart Pf/Pan test (100 and 96.15%, respectively). The CareStart Pf/Pan test also showed excellent agreement with the field microscopy results.

Conclusion

The Carestart Pf/Pan could be used as an alternative diagnostic test in malaria-endemic areas where facility for performing microscopy is not available.

Assessing Asymptomatic Malaria Carriage of Plasmodium falciparum and Non-falciparum Species in Children Resident in Nkolbisson, Yaoundé, Cameroon

Malaria is still a threat to public health as it remains the first endemic disease in the world. It is a pervasive parasitic disease in tropical and subtropical regions where asymptomatic malaria infection among humans serves as a significant reservoir for transmission. A rapid and correct diagnosis is considered to be an important strategy in the control of the disease especially in children, who are the most vulnerable group. This study assessed the prevalence of asymptomatic malaria in children at the Nkolbisson health area in Yaoundé, Cameroon. A cross-sectional study design and a convenience sampling plan were used. A total of 127 participants were recruited after informed and signed consent from parents and/or guardians. Blood samples were collected by finger-pricking and venipuncture from children aged 6 months to 10 years and then screened for asymptomatic parasitemia by a rapid diagnostic test (RDT), light microscopy (LM) staining with Giemsa and 18S rRNA polymerase chain reaction (PCR) for speciation. The data were analyzed using SPSS version 20 software. The study identified 85 children who were positive from the PCR, 95 positive from the RDT and 71 from the LM, revealing a malaria prevalence of 66.9%, 74.8% and 55.9%, respectively. The prevalence was not observed to be dependent on the sex and age group of the participants. Plasmodium falciparum was the predominant species followed by Plasmodium malariae and then Plasmodium ovale. The RDT and LM had the same sensitivity (90.6%) with a slight difference in their specificity (RDT: 57.1%; LM: 54.8%). The RDT also demonstrated higher positive and negative predictive values compared with those of the LM.

Diagnostic Accuracy of CareStart™ Malaria HRP2 and SD Bioline Pf/PAN for Malaria in Febrile Outpatients in Varying Malaria Transmission Settings in Cameroon

Background: There was an increase in the number of malaria cases in Cameroon in 2018 that could reflect changes in provider practice, despite effective interventions. In this study, we assessed the diagnostic performance of two malaria rapid diagnostic tests (mRDTs) for diagnostic confirmation of suspected cases of malaria in public and private health facilities in two malaria transmission settings in Cameroon. Methods: We evaluated the diagnostic performance of CareStart pf and SD Bioline Pf/PAN mRDT and compared these parameters by RDT type and transmission setting. Nested PCR and blood film microscopy were used as references. The chi square test was used for independent sample comparisons, while the McNemar’s test was used to test for the dependence of categorical data in paired sample testing. A p < 0.05 was considered significant in all comparisons. The R (v.4.0.2) software was used for analyses. Results: A total of 1126 participants consented for the study in the four sites. The diagnostic accuracy of the CareStart Pf mRDT was 0.93.6% (0.911–0.961) in Yaoundé, 0.930% (0.90–0.960) in Ngounso, 0.84% (0.794–0.891) in St Vincent Catholic Hospital Dschang and 0.407 (0.345–0.468) in Dschang district hospital. For SD Bioline Pf/PAN the accuracy was 0.759 (0.738–0.846) for St Vincent Catholic Hospital Dschang and 0.426 (0.372–0.496) for the Dschang district hospital. The accuracy was slightly lower in each case but not statistically different when PCR was considered as the reference. The likelihood ratios of the positive and negative tests were high in the high transmission settings of Yaoundé (10.99 (6.24–19.35)) and Ngounso (14.40 (7.89–26.28)) compared to the low transmission settings of Dschang (0.71 (0.37–1.37)) and St Vincent Catholic hospital (7.37 (4.32−12.59)). There was a high degree of agreement between the tests in Yaoundé (Cohen’s Kappa: 0.85 ± 0.05 (0.7–0.95)) and Ngounso (Cohen’s Kappa: 0.86 ± 0.05 (0.74, 0.97)) and moderate agreement in St Vincent hospital Dschang (k: 0.58 ± 0.06 (0.44–0.71)) and poor agreement in the District Hospital Dschang (Cohen’s Kappa: −0.11 ± 0.05 (−0.21–0.01)). The diagnostic indicators of the SD Bioline Pf/PAN were slightly better than for CareStart Pf mRDT in St Vincent Catholic hospital Dschang, irrespective of the reference test. Conclusions: Publicly procured malaria rapid diagnostic tests in Cameroon have maintained high accuracy (91–94%) in the clinical diagnosis of malaria in high malaria transmission regions of Cameroon, although they failed to reach WHO standards. We observed an exception in the low transmission region of Dschang, West region, where the accuracy tended to be lower and variable between facilities located in this town. These results underscore the importance of the routine monitoring of the quality and performance of malaria RDTs in diverse settings in malaria endemic areas.

Evaluating the dual reactivity on SD bioline malaria rapid diagnosis tests as a potential indicator of high parasitemia due to Plasmodium falciparum

The co-reactivity of the Plasmodium histidine-rich protein 2 (HRP2) and lactate dehydrogenase (pLDH) in malaria rapid diagnosis tests (mRDTs) as a potential indicator of high parasitemia linked to Plasmodium falciparum was evaluated in the reported study from Cameroon. The samples were screened for malaria using both mRDTs (SD bioline HRP2/pLDH), light microscopy and further confirmed by Plasmodium species-specific PCR assay. Of the 483 patients enrolled, 161 (33.3%) showed a reactive mRDTs amongst which 70 patients were positive by both microscopy and mRDTs with 30.0% (21/70) positive for HRP2 alone, while 70.0% (49/70) showed a dual reaction to HRP2 and pLDH parasite antigens. P. falciparum parasitemia was found to be significantly high among patients with both reactive antigens, (p < 0.0001) suggesting that mRDTs reactivity is influenced by parasite load which could be used as a diagnostic marker for therapeutic management of patients with high parasitemia in field conditions.

Comparison of conventional and non-invasive diagnostic tools for detecting Plasmodium falciparum infection in southwestern Cameroon: a cross-sectional study

Background

Malaria remains a significant health challenge in sub-Saharan Africa, with early diagnosis critical to reducing its morbidity and mortality. Despite the increasing Plasmodium spp. diagnostic capabilities, access to testing is limited in some cases by the almost absolute requirement for blood from potentially infected subjects as the only sample source for all conventional methods. A rapid test on non-invasive specimen with comparable performance to microscopy for the screening or diagnosis of all participants is invaluable. This study sought to compare conventional and non-invasive diagnostic tools for detecting Plasmodium falciparum.

Methods

This was a cross-sectional study, carried out between March and August 2019 to evaluate and compare the diagnostic performance of a PfHRP2/pLDH-based malaria rapid diagnostic test (mRDT) on patients’ blood, saliva and urine relative to conventional light microscopy and nested PCR at outpatient clinics in the Buea and Tiko health districts of Southwestern Cameroon. The significance of differences in proportions was explored using the Pearson’s χ² test whereas differences in group means were assessed using analyses of variance.

Results

A total of 359 individuals of both sexes, aged 1–92 years, were enrolled into the study. Of the 301 individuals tested by light microscopy and mRDTs on blood, saliva and urine, 84 (27.9%), 81 (26.9%), 87 (28.9%) and 107 (35.5%) respectively were positive. However, only 34.3%, 90.5%, 91.4%, 83.9% and 65.4% febrile, light microscopy and mRDT positives on blood, saliva and urine respectively had P. falciparum infection as confirmed by PCR. The sensitivity and specificity of presumptive diagnosis, light microscopy and mRDT on blood, saliva and urine were 86.9% and 19.7%, 77.8% and 96.1%, 75.8% and 96.6%, 74.5% and 93.1%, and 70.7% and 81.8%, respectively. The agreement between mRDT on saliva (k = 0.696) and microscopy (k = 0.766) compared to PCR was good.

Conclusion

The study highlighted the low performance of presumptive diagnosis, reinforcing the need for parasitological tests prior to antimalarial therapy. The higher PfHRP2/pLDH mRDT parasite detection rates and sensitivity in saliva compared to urine suggests that the former is a practical adjunct to or alternative worth optimising for the routine diagnosis of malaria.

Graphic Abstract

Flow chart for diagnosis of P. falciparum infection by light microscopy, rapid diagnostic tests and nested PCR.

Epidemiological and entomological studies of malaria transmission in Tibati, Adamawa region of Cameroon 6 years following the introduction of long-lasting insecticide nets

BACKGROUND

Malaria remains a serious public health problem in Cameroon. Implementation of control interventions requires prior knowledge of the local epidemiological situation. Here we report the results of epidemiological and entomological surveys carried out in Tibati, Adamawa Region, Cameroon, an area where malaria transmission is seasonal, 6 years after the introduction of long-lasting insecticidal bed nets.

METHODS

Cross-sectional studies were carried out in July 2015 and 2017 in Tibati. Thick blood smears and dried blood spots were collected from asymptomatic and symptomatic individuals in the community and at health centers, respectively, and used for the molecular diagnosis of Plasmodium species. Adult mosquitoes were collected by indoor residual spraying and identified morphologically and molecularly. The infection status of Plasmodium spp. was determined by quantitative PCR, and positivity of PCR-positive samples was confirmed by Sanger sequencing.

RESULTS

Overall malaria prevalence in our study population was 55.0% (752/1367) and Plasmodium falciparum was the most prevalent parasite species (94.3%), followed by P. malariae (17.7%) and P. ovale (0.8%); 92 (12.7%) infections were mixed infections. Infection parameters varied according to clinical status (symptomatic/asymptomatic) and age of the sampled population and the collection sites. Infection prevalence was higher in asymptomatic carriers (60.8%), but asexual and sexual parasite densities were lower. Prevalence and intensity of infection decreased with age in both the symptomatic and asymptomatic groups. Heterogeneity in infections was observed at the neighborhood level, revealing hotspots of transmission. Among the 592 Anopheles mosquitoes collected, 212 (35.8%) were An. gambiae, 172 (29.1%) were An. coluzzii and 208 (35.1%) were An. funestus (s.s.). A total of 26 (4.39%) mosquito specimens were infected by Plasmodium sp. and the three Anopheles mosquitoes transmitted Plasmodium at equal efficiency. Surprisingly, we found an An. coluzzii specimen infected by Plasmodium vivax, which confirms circulation of this species in Cameroon. The positivity of all 26 PCR-positive Plasmodium-infected mosquitoes was successively confirmed by sequencing analysis.

CONCLUSION

Our study presents the baseline malaria parasite burden in Tibati, Adamawa Region, Cameroon. Our results highlight the high malaria endemicity in the area, and hotspots of disease transmission are identified. Parasitological indices suggest low bednet usage and that implementation of control interventions in the area is needed to reduce malaria burden. We also report for the first time a mosquito vector with naturally acquired P. vivax infection in Cameroon.