1
|
Geldenhuys DS, Josias S, Brink W, Makhubele M, Hui C, Landi P, Bingham J, Hargrove J, Hazelbag MC. Deep learning approaches to landmark detection in tsetse wing images. PLoS Comput Biol 2023; 19:e1011194. [PMID: 37363914 PMCID: PMC10328335 DOI: 10.1371/journal.pcbi.1011194] [Citation(s) in RCA: 1] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/19/2022] [Revised: 07/07/2023] [Accepted: 05/17/2023] [Indexed: 06/28/2023] Open
Abstract
Morphometric analysis of wings has been suggested for identifying and controlling isolated populations of tsetse (Glossina spp), vectors of human and animal trypanosomiasis in Africa. Single-wing images were captured from an extensive data set of field-collected tsetse wings of species Glossina pallidipes and G. m. morsitans. Morphometric analysis required locating 11 anatomical landmarks on each wing. The manual location of landmarks is time-consuming, prone to error, and infeasible for large data sets. We developed a two-tier method using deep learning architectures to classify images and make accurate landmark predictions. The first tier used a classification convolutional neural network to remove most wings that were missing landmarks. The second tier provided landmark coordinates for the remaining wings. We compared direct coordinate regression using a convolutional neural network and segmentation using a fully convolutional network for the second tier. For the resulting landmark predictions, we evaluate shape bias using Procrustes analysis. We pay particular attention to consistent labelling to improve model performance. For an image size of 1024 × 1280, data augmentation reduced the mean pixel distance error from 8.3 (95% confidence interval [4.4,10.3]) to 5.34 (95% confidence interval [3.0,7.0]) for the regression model. For the segmentation model, data augmentation did not alter the mean pixel distance error of 3.43 (95% confidence interval [1.9,4.4]). Segmentation had a higher computational complexity and some large outliers. Both models showed minimal shape bias. We deployed the regression model on the complete unannotated data consisting of 14,354 pairs of wing images since this model had a lower computational cost and more stable predictions than the segmentation model. The resulting landmark data set was provided for future morphometric analysis. The methods we have developed could provide a starting point to studying the wings of other insect species. All the code used in this study has been written in Python and open sourced.
Collapse
Affiliation(s)
- Dylan S. Geldenhuys
- The South African Department of Science and Innovation-National Research Foundation (DSI-NRF) South African Centre for Epidemiological Modelling and Analysis (SACEMA), Stellenbosch University, Stellenbosch, South Africa
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
| | - Shane Josias
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
- School for Data Science and Computational Thinking, Stellenbosch University, Stellenbosch, South Africa
| | - Willie Brink
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
| | - Mulanga Makhubele
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
| | - Cang Hui
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
- Mathematical Biosciences Group, African Institute for Mathematical Sciences, Muizenberg, South Africa
| | - Pietro Landi
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
| | - Jeremy Bingham
- The South African Department of Science and Innovation-National Research Foundation (DSI-NRF) South African Centre for Epidemiological Modelling and Analysis (SACEMA), Stellenbosch University, Stellenbosch, South Africa
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
| | - John Hargrove
- The South African Department of Science and Innovation-National Research Foundation (DSI-NRF) South African Centre for Epidemiological Modelling and Analysis (SACEMA), Stellenbosch University, Stellenbosch, South Africa
- Department of Mathematical Sciences, Stellenbosch University, Stellenbosch, South Africa
| | - Marijn C. Hazelbag
- The South African Department of Science and Innovation-National Research Foundation (DSI-NRF) South African Centre for Epidemiological Modelling and Analysis (SACEMA), Stellenbosch University, Stellenbosch, South Africa
- ExploreAI (Pty) Ltd., Cape Town, South Africa
| |
Collapse
|
2
|
Mbewe NJ, Sole CL, Pirk CWW, Masiga DK, Yusuf AA. Efficiencies of stationary sampling tools for the tsetse fly Glossina fuscipes fuscipes in western Kenya. Acta Trop 2021; 223:106092. [PMID: 34389328 DOI: 10.1016/j.actatropica.2021.106092] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/30/2021] [Revised: 08/01/2021] [Accepted: 08/01/2021] [Indexed: 10/20/2022]
Abstract
Monitoring the effectiveness of tsetse fly control interventions that aim to reduce transmission of African trypanosomiasis requires highly efficient sampling tools that can catch flies at low densities. The sticky small target (StS-target) has previously been shown to be more effective in sampling Glossina fuscipes fuscipes compared to the biconical trap. However, its efficiency in terms of the proportion of flies it catches out of those that visit it has not been reported. Furthermore, there are no reports on whether tsetse samples caught using the StS-target can be used for subsequent processes such as molecular tests. In this study, we evaluated the efficiency of the biconical trap and targets for sampling G. f. fuscipes. All targets were tiny (0.25 × 0.50 m) but varied in their capture system. We used targets with sticky surface (StS-targets) and those with an electrified surface (ES-targets). We also assessed the suitability of flies caught on the StS-target for molecular tests by amplifying DNA of bacterial communities. Randomized block design experiments were undertaken in Mbita area and Manga Island on Lake Victoria of western Kenya. Fly catches of each sampling tool were compared to those of the sampling tool flanked by electric (E) nets and analyzed using a negative binomial regression. The total catch for each sampling tool alone was divided by the total catch of the sampling tool flanked by two E-nets to obtain its efficiency expressed as a percentage. A proportion of flies caught on the StS-target was preserved for molecular tests. Overall, the efficiencies of the biconical trap, ES-target and StS-target were 7.7%, 13.3% and 27.0%, respectively. A higher proportion of females (69 to 79%) than males approached all the sampling tools, but the trap efficiency was greater for male G. f. fuscipes than females. Furthermore, sequencing the 16S rRNA gene from fly samples caught on the StS-target revealed the presence of Spiroplasma. Our results indicate that the SS-target is the most efficient trap to monitor G. f. fuscipes population during interventions, with the biconical trap being the least efficient, and samples collected from StS-targets are suitable for molecular studies.
Collapse
|
3
|
Mweempwa C, Mbewe NJ, De Deken R. Wing length of tsetse caught by stationary and mobile sampling methods. Acta Trop 2020; 204:105333. [PMID: 31926912 DOI: 10.1016/j.actatropica.2020.105333] [Citation(s) in RCA: 3] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/03/2019] [Revised: 01/02/2020] [Accepted: 01/05/2020] [Indexed: 10/25/2022]
Abstract
INTRODUCTION A variety of techniques have been used to control tsetse with varying degrees of success. In a study on the population structure of Glossina fuscipes fuscipes that recovered after a previous vector control trial on two Kenyan islands, it was reported that the average fly size on the intervention islands was significantly smaller than on the none intervention islands and also compared to the size before the intervention. The conclusion was that vector control using tiny targets exerted size selection pressure on the population. The study recommended for further studies and suggested that this phenomenon could be among the reasons why targets used as a sole control method have rare reports of successful elimination of tsetse populations. Therefore, in this paper we report on a study of body size of tsetse flies caught in epsilon traps (as a stationary device) and black screen fly rounds (as a mobile trapping device). MATERIALS AND METHODS The study was carried out in eastern Zambia to test the hypothesis that the body size (measured as wing length) of G. m. morsitans males or females, captured by epsilon traps and fly rounds is the same. RESULTS A total of 1442 (489 females and 953 males) wing length measurements of G. m. morsitans were used in the analysis. It was established that tsetse flies caught by epsilon traps are on average larger than those caught by fly rounds. The likelihood of a large female or male fly being caught by traps, relative to a small one, significantly increased by 5.088 times (95% CI: 3.138-8.429) and by 2.563 times (95% CI: 1.584-4.148), respectively, p < 0.0001, compared with being caught by fly rounds. The hypothesis was rejected. CONCLUSION This study showed that epsilon traps capture significantly larger G. m. morsitans than fly rounds do. Therefore, further research is recommended to verify i) whether the predilection of traps to capture larger flies has an effect on the process of tsetse elimination when targets are used e.g. targets may take longer to reach elimination than if the predilection was not there, ii) whether different results can be obtained on ecogeographic distribution of different sizes of the species if fly rounds are used for sampling instead of epsilon traps. The results from such studies could influence the strategies used in future control operations.
Collapse
|
4
|
Mbewe NJ, Saini RK, Irungu J, Yusuf AA, Pirk CWW, Torto B. Responses of Glossina fuscipes fuscipes to visually attractive stationary devices baited with 4-methylguaiacol and certain repellent compounds in waterbuck odour. PLoS Negl Trop Dis 2019; 13:e0007510. [PMID: 31276492 PMCID: PMC6636772 DOI: 10.1371/journal.pntd.0007510] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/11/2019] [Revised: 07/17/2019] [Accepted: 06/03/2019] [Indexed: 11/18/2022] Open
Abstract
BACKGROUND A blend of compounds (pentanoic acid, guaiacol, δ-octalactone and geranylacetone) identified in waterbuck (Kobus defassa) body odour referred to as waterbuck repellent compounds (WRC) and a synthetic repellent 4-methylguaiacol have previously been shown to repel tsetse flies from the morsitans group. However, these repellents have not been evaluated on palpalis group tsetse flies. In this study, we evaluated the effect of these repellents on catches of Glossina fuscipes fuscipes (major vector of human sleeping sickness) in biconical traps and on sticky small targets which are visually attractive to palpalis group flies. The attractive devices were baited with different doses and blends of the repellent compounds. We also assessed the effect of removal of individual constituents in the synthetic blend of WRC on catches of G. f. fuscipes. METHODOLOGY/PRINCIPAL FINDINGS The study was conducted in western Kenya on four islands of Lake Victoria namely Big Chamaunga, Small Chamaunga, Manga and Rusinga. The tsetse fly catches from the treatments were modeled using a negative binomial regression to determine their effect on catches. In the presence of WRC and 4-methylguaiacol (released at ≈2 mg/h and ≈1.4 mg/h respectively), catches of G. f. fuscipes were significantly reduced by 33% (P<0.001) and 22% (P<0.001) respectively in biconical traps relative to control. On sticky small targets the reduction in fly catches were approximately 30% (P<0.001) for both 4-methylguiacol and WRC. In subtractive assays, only removal of geranylacetone from WRC significantly increased catches (by 1.8 times; P <0.001) compared to the complete blend of WRC. CONCLUSIONS/SIGNIFICANCE We conclude that WRC and 4-methylguaiacol reduce catches of G. f. fuscipes at stationary visually attractive traps and suggest that they may serve as broad spectrum repellents for Glossina species. We recommend further studies to investigate the effects of these compounds on reduction of G. f. fuscipes attracted to human hosts as this may lead to development of new strategies of reducing the prevalence and incidence of sleeping sickness.
Collapse
Affiliation(s)
- Njelembo J. Mbewe
- International Centre of Insect Physiology and Ecology, Nairobi, Kenya
- Department of Zoology and Entomology, University of Pretoria, Hatfield, Pretoria, South Africa
| | - Rajinder K. Saini
- International Centre of Insect Physiology and Ecology, Nairobi, Kenya
- Pestinix, International Pest & Vector Control Specialists, Nairobi, Kenya
| | - Janet Irungu
- International Centre of Insect Physiology and Ecology, Nairobi, Kenya
| | - Abdullahi A. Yusuf
- Department of Zoology and Entomology, University of Pretoria, Hatfield, Pretoria, South Africa
| | - Christian W. W. Pirk
- Department of Zoology and Entomology, University of Pretoria, Hatfield, Pretoria, South Africa
| | - Baldwyn Torto
- International Centre of Insect Physiology and Ecology, Nairobi, Kenya
- Department of Zoology and Entomology, University of Pretoria, Hatfield, Pretoria, South Africa
| |
Collapse
|
5
|
Improved estimates for extinction probabilities and times to extinction for populations of tsetse (Glossina spp). PLoS Negl Trop Dis 2019; 13:e0006973. [PMID: 30964873 PMCID: PMC6474634 DOI: 10.1371/journal.pntd.0006973] [Citation(s) in RCA: 9] [Impact Index Per Article: 1.8] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/31/2018] [Revised: 04/19/2019] [Accepted: 02/04/2019] [Indexed: 11/19/2022] Open
Abstract
A published study used a stochastic branching process to derive equations for the mean and variance of the probability of, and time to, extinction in population of tsetse flies (Glossina spp) as a function of adult and pupal mortality, and the probabilities that a female is inseminated by a fertile male. The original derivation was partially heuristic and provided no proofs for inductive results. We provide these proofs, together with a more compact way of reaching the same results. We also show that, while the published equations hold good for the case where tsetse produce male and female offspring in equal proportion, a different solution is required for the more general case where the probability (β) that an offspring is female lies anywhere in the interval (0, 1). We confirm previous results obtained for the special case where β = 0.5 and show that extinction probability is at a minimum for β > 0.5 by an amount that increases with increasing adult female mortality. Sensitivity analysis showed that the extinction probability was affected most by changes in adult female mortality, followed by the rate of production of pupae. Because females only produce a single offspring approximately every 10 days, imposing a death rate of greater than about 3.5% per day will ensure the eradication of any tsetse population. These mortality levels can be achieved for some species using insecticide-treated targets or cattle—providing thereby a simple, effective and cost-effective method of controlling and eradicating tsetse, and also human and animal trypanosomiasis. Our results are of further interest in the modern situation where increases in temperature are seeing the real possibility that tsetse will go extinct in some areas, without the need for intervention, but have an increased chance of surviving in other areas where they were previously unsustainable due to low temperatures. We derive equations for the mean and variance of the probability of, and time to, extinction in population of tsetse flies (Glossina spp), the vectors of trypanosomiasis in sub-Saharan Africa. In so doing we provide the complete proofs for all results, which were not provided in a previously published study. We also generalise the derivation to allow for the probability that an offspring is female to lie anywhere in the interval (0, 1). The probability of extinction was most sensitive to changes in adult female mortality. The unusual tsetse life cycle, with very low reproductive rates, means that populations can be eradicated as long as adult female mortality is raised to levels greater than about 3.5% per day. Simple bait methods of tsetse control, such as insecticide-treated targets and cattle, can therefore provide simple, affordable and effective means of eradicating tsetse populations. The results are of further interest in the modern situation where increases in temperature are seeing the real possibility that tsetse will go extinct in some areas, but have an increased chance of surviving in others where they were previously unsustainable due to low temperatures.
Collapse
|
6
|
Hargrove J, English S, Torr SJ, Lord J, Haines LR, van Schalkwyk C, Patterson J, Vale G. Wing length and host location in tsetse (Glossina spp.): implications for control using stationary baits. Parasit Vectors 2019; 12:24. [PMID: 30635017 PMCID: PMC6329045 DOI: 10.1186/s13071-018-3274-x] [Citation(s) in RCA: 10] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/26/2018] [Accepted: 12/19/2018] [Indexed: 11/10/2022] Open
Abstract
BACKGROUND It has been suggested that attempts to eradicate populations of tsetse (Glossina spp.) using stationary targets might fail because smaller, less mobile individuals are unlikely to be killed by the targets. If true, tsetse caught in stationary traps should be larger than those from mobile baits, which require less mobility on the part of the flies. RESULTS Sampling tsetse in the Zambezi Valley of Zimbabwe, we found that the number of tsetse caught from stationary traps, as a percent of total numbers from traps plus a mobile vehicle, was ~5% for male G. morsitans morsitans (mean wing length 5.830 mm; 95% CI: 5.800-5.859 mm) and ~10% for females (6.334 mm; 95% CI: 6.329-6.338 mm); for G. pallidipes the figures were ~50% for males (6.830 mm; 95% CI: 6.821-6.838 mm) and ~75% for females (7.303 mm, 95% CI: 7.302-7.305 mm). As expected, flies of the smaller species (and the smaller sex) were less likely to be captured using stationary, rather than mobile sampling devices. For flies of a given sex and species the situation was more complex. Multivariable analysis showed that, for females of both species, wing lengths changed with ovarian age and the month, year and method of capture. For G. pallidipes, there were statistically significant interactions between ovarian age and capture month, year and method. For G. m. morsitans, there was only a significant interaction between ovarian age and capture month. The effect of capture method was, however, small in absolute terms: for G. pallidipes and G. m. morsitans flies caught on the mobile vehicle had wings only 0.24 and 0.48% shorter, respectively, than flies caught in stationary traps. In summary, wing length in field samples of tsetse varies with ovarian age, capture month and year and, weakly, with capture method. Suggestions that a target-based operation against G. f. fuscipes in Kenya caused a shift towards a smaller, less mobile population of tsetse, unavailable to the targets, failed to account for factors other than capture method. CONCLUSIONS The results are consistent with the successful use of targets to eradicate populations of tsetse in Zimbabwe. Until further, more nuanced, studies are conducted, it is premature to conclude that targets alone could not, similarly, be used to eradicate G. f. fuscipes populations in Kenya.
Collapse
Affiliation(s)
- John Hargrove
- SACEMA, University of Stellenbosch, Stellenbosch, South Africa.
| | - Sinead English
- School of Biological Sciences, University of Bristol, Bristol, UK
| | | | - Jennifer Lord
- Liverpool School of Tropical Medicine, Liverpool, UK
| | | | | | | | - Glyn Vale
- SACEMA, University of Stellenbosch, Stellenbosch, South Africa.,Natural Resources Institute, University of Greenwich, London, UK
| |
Collapse
|
7
|
Mbewe NJ, Saini RK, Torto B, Irungu J, Yusuf AA, Pirk CWW. Sticky small target: an effective sampling tool for tsetse fly Glossina fuscipes fuscipes Newstead, 1910. Parasit Vectors 2018; 11:268. [PMID: 29695261 PMCID: PMC5922315 DOI: 10.1186/s13071-018-2840-6] [Citation(s) in RCA: 5] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/28/2017] [Accepted: 04/09/2018] [Indexed: 12/11/2022] Open
Abstract
Background Small targets comprising panels of blue and insecticide-treated black netting material each 0.25 × 0.25 m have been shown to attract and kill Glossina fuscipes fuscipes Newstead, 1910 (Diptera: Glossinidae) thereby reducing its population density by over 90% in field trials. However, their attractive ability has not been fully exploited for sampling purposes. Therefore, in this study we assessed the effectiveness of using sticky small targets as sampling tools for G. f. fuscipes in western Kenya. We also determined the influence of colour on the landing response of female and male flies on sticky small targets. Methods Using a series of randomised block experiments, the numbers of tsetse flies caught with sticky small targets were compared with those caught with biconical traps. A negative binomial regression was used to model fly catches. Odds ratios as measures of association between the landing response on the blue or black panel of the sticky small target and the sex of flies were obtained from a multiple logistic regression. Results The results showed that sticky small targets caught 13.5 and 3.6 times more female and male tsetse flies than biconical traps (Z = 9.551, P < 0.0001 and Z = 5.978, P < 0.0001, respectively). Females had a 1.7 times likelihood of landing on the black panel than males (Z = 2.25, P = 0.025). Conclusion This study suggests that sticky small targets are an effective sampling tool for G. f. fuscipes. Therefore, we recommend the use of sticky small targets as an alternative to biconical traps for observational and experimental investigations of G. f. fuscipes.
Collapse
Affiliation(s)
- Njelembo J Mbewe
- International Centre of Insect Physiology and Ecology, P.O Box 30772-00100, Nairobi, Kenya. .,Department of Zoology and Entomology, University of Pretoria, Private Bag X20, Hatfield, Pretoria, 028, South Africa.
| | - Rajinder K Saini
- International Centre of Insect Physiology and Ecology, P.O Box 30772-00100, Nairobi, Kenya.,Pestinix, International Pest & Vector Control Specialists, P.O. Box 702-00621, Nairobi, Kenya
| | - Baldwyn Torto
- International Centre of Insect Physiology and Ecology, P.O Box 30772-00100, Nairobi, Kenya.,Department of Zoology and Entomology, University of Pretoria, Private Bag X20, Hatfield, Pretoria, 028, South Africa
| | - Janet Irungu
- International Centre of Insect Physiology and Ecology, P.O Box 30772-00100, Nairobi, Kenya
| | - Abdullahi A Yusuf
- Department of Zoology and Entomology, University of Pretoria, Private Bag X20, Hatfield, Pretoria, 028, South Africa
| | - Christian W W Pirk
- Department of Zoology and Entomology, University of Pretoria, Private Bag X20, Hatfield, Pretoria, 028, South Africa
| |
Collapse
|