Using Imaging Technologies to Control Malaria
Benjamin G. Jacob
Jose E. Funes
Ephantus J. Muturi
Robert J. Novak
Illinois Natural History Survey
Center for Ecological Entomology
James L. Regens
Center for Biosecurity Research
University of Oklahoma Health Sciences Center
Oklahoma City, Okla.
Josephat I. Shililu
John I. Githure
Human Health Division
International Centre of Insect Physiology and Ecology (ICIPE)
Malaria, virtually unknown in developed countries, exacts a horrific toll in terms of human health throughout the developing regions of the world. One to three million deaths occur from the disease each year, and most of the global burden falls on sub-Saharan Africa.
To place the impact on sub-Saharan Africa in perspective, 300 to 500 million cases of malaria are reported annually there – more than the total number of people living in the United States. In some areas people receive 200 to 300 infective bites per year, and those who die are overwhelmingly children under five years of age.
Efforts during the 20th century demonstrated that reducing the abundance of larval habitats for Anopheline mosquitoes is a key tool for malaria control.
Because suppressing mosquito densities is one of the critical factors in reducing the transmission of the malaria parasite, identification of the distribution of viable habitats is important.
|Figure 2 QuickBird 0.61-m spatial resolution of urban and vegetation land cover in Malindi, Kenya|
|Figure 3 IKONOS 4-m spatial resolution of Kangichiri agro-village complex in the Mwea Rice Scheme, Kenya|
Remote sensing (RS) imaging technologies offer cost-effective means for identifying habitats, estimating densities of mosquito species, and predicting disease incidence to support vector control programs. Typically, multispectral information from visible and near-infrared (NIR) light wavelengths covering a range from 0.45 to 0.96 µm can distinguish between high and low mosquito-producing areas. These, in turn, are correlated with vector distributions as well as malaria incidence and prevalence. Imaging provides an effective way to identify likely mosquito habitats for implementing various strategies to control the spread of malaria.
MAPPING HABITATS WITH IMAGERY DATA
Most studies have used data from Landsat or the National Oceanic and Atmospheric Administration's (NOAA) Advanced Very High Resolution Radiometer (AVHRR), although those systems have limited spatial resolution for characterizing and monitoring the spatial and temporal patterns of mosquito habitats (see Figure 1). In recent years, however, there has been a significant advance in high-resolution polarimetric satellite platforms.
The best-known of the fine spatial resolution satellite sensors, and regarded widely as the first of this new “era” of remote sensing, is IKONOS, launched in 1999 (GeoEye, Dulles, Va.). The IKONOS satellite uses a multispectral sensor to collect blue, green, red, and NIR bands with 4.0-m resolution, providing natural-color imagery for visual interpretation and color-infrared applications. The IKONOS satellite also collects panchromatic imagery at 1-m resolution.
The launch of the QuickBird satellite in 2001 provided even more refined spatial resolution imagery (0.6-m and 2.4-m spatial resolution panchromatic and multispectral imagery, respectively). IKONOS and QuickBird imagery can be used to generate geographic coverages (representative shapes) at a spatial scale equivalent to larval habitats (see Figures 2 and 3).
Employing geographic information systems (GIS) to digitally overlay higher spatial resolution satellite data with older, lower resolution Thematic Mapper data spanning back several decades offers a way to fuse data from multiple platforms.
Data fusion with GIS makes possible systematic delineation of seasonal entomologic habitats using remotely sensed data. Thus, it is possible to map significant land use variation, and to identify factors that can affect mosquito larval ecology (Figure 4). GIS can also provide relevant information that focuses control on the immature stages of vector Anopheles species to reduce the transmission of malaria.
|Figure 4 Land use land cover change (LULC) and non-LULC change areas mapped in GIS using July 1988 Thematic Mapper 30-m data and July 2005 IKONOS 4.0-m for Kangichiri village in the Mwea Rice Scheme, Kenya|
|Figure 5 NDVI using QuickBird red band and NIR data for Kangichiri village in the Mwea Rice Scheme, Kenya|
Climatic factors associated with malaria risks in sub-Saharan Africa can also be identified using meteorological data obtained from satellites along with malaria transmission distribution maps. Malaria transmission maps can be developed by taking into account biological constraints of climate on urban and rice land Anopheline habitat development. Widely used environmental proxies for determining the occurrence and distribution of Anopheline aquatic habitats include vegetation indices such as the normalized difference vegetation index (NDVI). NDVI expresses the abundance of actively photosynthesizing vegetation, or “greenness,” and has been of particular interest in mapping both spatial and temporal relationships between the environment and disease incidence (Figure 5).
The Image Analysis extension of Arc-View 9.1 was used to perform the NDVI calculations using ERDAS Imagine and ENVI software. ERDAS Imagine and ENVI are broad collections of software tools designed specifically to process satellite imagery. NDVI is calculated using the red channel (band 3) and the NIR channel (band 4). The NDVI is calculated from these band measurements as follows: (NIR band - red band) / (NIR band + red band). The NDVI calculation provides index values ranging from -1 to 1. NDVI values typically range from 0.1 up to 0.6 for vegetation, with higher values associated with greater density and greenness of the plant canopy. Surrounding soil and rock values are close to zero while the index values for water bodies such as rivers and dams have the opposite trend to vegetation and are negative.
SAMPLING USING SATELLITE IMAGERY
Both increased computing power and spatial modeling capabilities of GIS extend the use of remote sensing beyond the research community into operational disease surveillance and control. For example, overlaying a GIS sampling scheme on high resolution remotely sensed data helps organize and characterize mosquito larval habitats. A GIS sampling scheme is constructed by applying a mathematical algorithm to generate geometric cells (square, hexagonal, triangular) with a continuous and bounded surface consisting of equidistant cells that define an area to be sampled in the field to quantify attributes of interest. For instance, cells may be selected from the GIS sampling scheme in which the number, type, and area of larval habitats may be measured, or the number of larva and the proportion infected with the malaria parasite may be quantified.
GIS sampling scheme data files consist of columns and rows of uniform cells coded according to data values. Each cell within a matrix contains an attribute value as well as location coordinates and can be joined relationally to other databases. The spatial location of each cell may be implicitly contained within the ordering of the matrix (see Figure 6).
|Figure 6 A 250-m square urban grid for Malindi, Kenya|
|Figure 7 Digitized grid with a unique identifier with each grid cell/paddy overlaid on Karima village complex in the Mwea Rice Scheme|
Digitally tracing a rice land habitat in ArcInfo 9.1 generates polygons which can conform to rice land Anopheline aquatic habitat boundaries. The larval habitats can then be characterized in relation to ecological attributes of an aquatic habitat such as water temperature, flow, turbidity, or acidity (see Figure 7). Like the sampling scheme data files described earlier, each polygon is assigned an identifying number and the attributes sampled at each rice field can be added to the database.
INTEGRATING GPS TECHNOLOGY
One critical aspect that is often overlooked is the accuracy of global positioning system (GPS) coordinates. Accuracy in GPS/GIS is the degree of conformance between the estimated or measured position, time, and/or velocity of a GPS receiver and its true time, position, and/or velocity as compared with a constant standard. Developing an integrated vector management (IVM) program for control interventions requires relatively precise knowledge of the geographic location and ecological characteristics of habitats.
Members of this team recently compared the accuracy of location data collected from five handheld units (Trimble Geoexplorer 3, Garmin GPSMAP 76, Magellan 2000 XL, Garmin GPS 2PLUS and a highly precise CSI Wireless Max receiver) using real-time differentially corrected OmniStar L-Band satellite signal at the Illinois Natural History Survey (Champaign, Ill.). The coordinates generated from each of these were compared to five geodetic markers and the positional accuracy for each unit was determined. A histogram showing the distribution of the number of points in relation to the distance from the actual test location is shown in Figure 8.
The real-time CSI Wireless differentially-corrected global positioning system (DGPS) Max receiver, using real-time Omni Star L-Band satellite signal, mapped the geodetic sites with greater repeatable positional accuracy and higher distance accuracy than the other units. The repeatable positional accuracy improved by an average of 37.5 meters, and distance accuracy improved by an average of 31.5 meters through employing newer, more advanced CSI Wireless DGPS unit.
|Figure 8 The CSI Wireless Max receiver had the best positional accuracy. A total of 15 test points were collected. The Garmin 2 Plus had a positional accuracy of 15.0m (+/- 5.39m); Magellan 2000XL = 12.03m (+/- 4.17m); Trimble GeoExplorer 3 = 8.03m (+/- 2.43m); Garmin Map 76 = 3.34m (+/- .989m); the CSI Max receiver = .179 (+/- .392).|
Terrestrial DGPS and public access systems improves GPS position accuracy but are dependent on range and transmission interference and are proven to be less reliable than satellite-based systems. The CSI Wireless DGPS Max is a high-accuracy GPS receiver that incorporates internal sensors capable of receiving corrections from Space Based Augmentation Systems (SBAS), the worldwide OmniSTAR service, and DGPS beacon stations. OmniSTAR operates a network of 19 base stations that constantly communicate with available GPS satellites and calculate correction values. The OmniStar subscription service has the benefit of supplying fast and accurate GPS coordinate data without the added steps, inconvenience and processing time associated with other GPS acquisition systems. When using any of these services, the DGPS CSI Max receiver provides locational accuracy at position update rates of up to five times per second (5 Hz). With the growing demand for accurate and reliable GPS positioning, there has been a significant move towards the use of real-time GPS augmentation systems with wide area differential positioning capabilities for ecologically based field surveillance.
CREATING EFFECTIVE CONTROL PROGRAMS
Because malaria is still increasing in sub-Saharan Africa, it is crucial to identify accurate site-specific information to determine areas which maintain the zoonotic cycle. Determining remote heterogeneity in larval habitat distribution can have important operational significance. Evaluating the spatial-temporal distribution of larval habitats by integrating high-resolution satellite data and highly accurate ground coordinates in malaria-prone areas provides the foundation for implementing control measures based on habitat productivity.
Base maps were created from IKONOS and QuickBird data using ArcInfo 9.1 from DGPS ground coordinates. Each aquatic habitat, with associated land cover attributes from a study site, can be entered into a Vector Control Management System (VCMS) database, Advanced Computer Resources Corp. (ACR, Nashua, N.H.). VCMS can support mobile field data acquisition in a study site through a Microsoft PocketPC. All two-way, remote synchronizing of mosquito data, geocoding, and spatial display can be processed using the embedded GIS Interface Kit, which can be built using ESRI's MapOb-jects 2 technology. The VCMS database can plot and update DGPS ground coordinates of urban and rice land Anopheline aquatic larval habitat seasonal information and can support exporting data to a spatial format in which any combination of larval habitats and supporting data are exported in GIS as a shapefile. Such information makes possible the design and implementation of vector control operations that target zones where high larval densities occur (Figure 9).
Figure 9 Spatially targeting An.arabiensis aquatic habitat using a digitized grid algorithm and QuickBird visible and NIR data within a 1-km buffer divided by quadrants in Kangichiri agro-village complex in the Mwea Rice Scheme, Kenya
The current generation of imaging technology was applied to assess the effectiveness of control measures, including using new formulations of insecticides, at representative sites. Time series imagery acquired from field sites combined with laboratory analysis of Bacillus thuringiensis ssp. israelensis (Bti), Bacillus sphaericus (Neide) (Bsph), and Bti:Bsph ratios can be used for preliminary determination of lethal dose (LD) parameters. Follow-up confirmation of LD50 and LD95 values using village-scale tests for final candidate formulations can evaluate efficacy (based on feeding behavior and susceptibility to bacteria toxins), impact of ultraviolet radiation on efficacy, ease of use through conventional application equipment, and cost profile relative to other larvicides. Imagery data also can be used to assess the impact of any control measure on agricultural productivity.
The evolving state-of-the-art in imaging technologies and the expanding scope for applications offer tremendous promise for controlling malaria. These technologies can help transform the vision of eradicating malaria as a major public health threat into a reality. Furthermore, the techniques summarized here can also be applied to other vector-borne infectious diseases to promote other advances in public health.
For their data collection efforts and for conducting this study, the authors would like to thank the ICIPE Mwea Rice Mosquito Team: Simon M. Muriu, Enock Mpanga, James Wauna, Peter Barasa, Nelson M. Muchiri, Gladys Kamari, Irene Kamau, Charles C. Kiura, Peter M. Mutiga, Paul K. Mwangi, Nicholus G. Kamari, William M. Waweru, Christine W. Maina, Martin Njigoya and Naftaly Gichuki at the Mwea Divison in Kenya. This research was funded by the National Institute of Health Grant U01A1054889 (Robert Novak) University of Illinois, Urbana-Champaign, Ill.