Conceived and designed the experiments: CM JS JG JM RS. Performed the experiments: CM JS. Analyzed the data: CM JS. Contributed reagents/materials/analysis tools: KCL JG. Wrote the paper: CM JS.
The authors have declared that no competing interests exist.
Lymphatic filariasis is caused by filarial nematode parasites, including
The World Health Organization estimates that there are approximately 37 million people who are afflicted by
Lymphatic filariasis is a devastating parasitic disease that affects more than 120 million people in 81 countries
Lymphatic filariasis is spread mainly by three species of nematodes in the family Filariodidea:
The WHO currently advocates interrupting transmission of the disease via an annual mass drug administration of single doses of albendazole and either diethylcarbamazine or ivermectin. These drugs are effective at killing microfilariae but not effective against adult worms (macrofilariae). Since adult worms can live for up to 6–8 years, treatment must be given on a regular basis to break the cycle of transmission. This widespread treatment is logistically challenging and costly, particularly in endemic regions that are politically unstable. Such widespread application also raises the threat of resistance, whose first signs are being seen with ivermectin in
Currently there is no high throughput screening (HTS) method available to screen compounds targeting any of these macroscopic nematodes in vitro. Assays have been developed in recent years that score worm migration, feeding and development
We have developed an inexpensive system (“WormAssay”) for quantifying parasite movement based on worm motility. The apparatus uses a commodity video camera, computer and a newly developed free and open source software application to provide quantitative measurements of parasite motility on entire plates. The application can process multiple wells simultaneously without user interaction, and automatically identifies each well in the plate and labels the output data accordingly. This system can be used to assay large parasites such as the filarid nematodes as well as other macroparasites. WormAssay's automation of the video capture step and lack of need for any interaction with the computer software during scoring differentiates it from all existing motion-based schemes, and permits current screening of 400 worms per assay each week.
Individual adult
The assays were performed using the open source computer software program described here. This program is named WormAssay. Plates were visualized using a Canon HV-40 Vixia HDV camcorder (Canon Inc.) providing 1080p H.262/MPEG-2 Part 2 compressed HDV video connected via IEEE1384 to an Apple iMac with a 2.93 GHz Intel Core i7 4-core CPU (Apple Inc.) and WormAssay (version 0.15) for 1 minute. The application and source code are available for free use, modification and redistribution under the terms of the GNU Public License (version 2 or later; see
The WormAssay was developed for use in high throughput screening of
The video camera is positioned below the microtiter plate and the plate is recorded using the WormAssay software.
(A) Video frames of 24-well plates of
WormAssay's data acquisition does not require user interaction or configuration and is suitable for robotics integration with any multi-degree-of-freedom plate manipulator. Data acquisition automatically begins when a plate appears in the field of view of the camera and data is written immediately upon removal of the plate (Videos S4, S5 and S6). Video recordings of each read are archived. Motility and other assay data are written to CSV (spreadsheet style) files for use with standard statistical analysis software tools. Barcode reading is performed on the video stream (or from another video camera attached) to automatically label results. The application can automatically email results at the end of a run, for example, when used in a unattended automated assay.
An extended finite state machine
The nodes indicate distinct states (or the start/end pseudostates) and the edges indicate transition conditions.
The application has no knowledge of the specific geometry of the microtiter plates, except for the number of wells in each row and column of the supported plate sizes. This allows for great tolerance in terms of specific plate geometry and in the position of the plate within the camera's field of view. This is contrary to the scheme used by most microtiter plate assay equipment, where a mechanical sensor is positioned over or in the well of interest, one well at a time.
The WormAssay well finding algorithm iterates through acceptable plate configurations in parallel, corresponding to 6-, 12-, 24-, 48- and 96-well microtiter plates. First, Canny's algorithm
(A) Unprocessed video frame in a 96-well plate. (B) Screen capture of the software's user interface for those plates using the Consensus Voting Luminance Difference algorithm (see caption for
We developed two analysis algorithms. The first determines the average velocity of the moving contours inside each well. This algorithm derives the velocity from the optical flow vectors of the luminance component of the video stream from a pair of adjacent frames approximately 100 ms apart. The algorithm uses the sparse iterative version of the Lucas-Kanade optical flow in pyramids provided with the OpenCV framework
This algorithm is useful for scoring rates of motion (or lack thereof) of single parasites with high accuracy as it can reliably differentiate small differences in velocity which may correspond to differing amounts of motility inhibition. A velocity in single dimensional pixel units per second is reported. Only moving components are considered, so this assay is not suitable for assays where a combination of dead (motionless) and moving parasites are present in a single well, since only the moving parasites will be considered in the score. This algorithm is described in WormAssay's Options-Analyzer menu as “Lucas-Kanade Optical Flow.”
The second algorithm is an algorithm that detects changes in the occupation and vacancy of pixels between a group of frames. It uses difference information between a subset of 5 frames chosen at random from the frames that arrived in the past second. First, a difference is performed on each of the 3 color channels of each of the 5 frames and the current frame. Then high frequency components are removed from each of the set of 5 difference values. A voting scheme is employed to determine when a pixel has had its contents changed. Three or more changed pixels is deemed a quorum, otherwise the changes are ignored and deemed noise. The number of filled or vacated pixels is then summed and taken as a fraction of the total number of pixels within the well's circle (times 1000 to improve numerical readability). This number is reported as an arbitrary area unit indicating motility. This algorithm is useful for detecting very low levels of movement or for quantifying the aggregate movement of more than one parasite in a given well. This algorithm is described in the application as “Consensus Voting Luminance Difference.”
All algorithms process in real-time, in parallel on each well. The algorithm programming model is extensible; new algorithms can be added independently of other components of the application. To avoid recording spurious values when the plates are being moved at the beginning or end of a run, the software ignores any frames whose total motion (via the pixelwise mean of the simple interframe absolute difference across all color channels) exceeds a threshold. This value may need to be modified for assays with very large or motile organisms. This is the only non-general threshold used in the application. Improving this aberrant (whole plate) motion detection is a possible area of further research.
Both algorithms (and the well detection) are computationally intensive, and are not able to process every frame of the 1080p (1080×1920 pixels) video input, which is typically 24 or 30 frames per second. On a modern (2011) typical multicore desktop computer, we are able to process 5–10 frames per second, which yields satisfactory results. Since recording of all wells is done in parallel, this is significantly faster than the 5–10 minute recoding times necessary to generate even short 10 frames per second movies on a well-by-well basis on commercial plate microscopes (e.g. on the GE IN Cell Analyzer 2000.)
We also developed a dark-field parallel macroscopic imaging apparatus connected to an HDV camera with an IEEE1394 interface using inexpensive materials
The apparatus used consists of a light-tight box with a hinged lid on the top, with the video camera mounted outside (to ensure easy access and proper cooling) at the bottom of the box and recording upwards. The whole box is made of plywood with some metal parts, all painted black to minimize reflections. The plate is positioned above the camera at such a distance that allows the plate image to fully fill the field of view (approx. 35 cm). The plate is illuminated by a dimmable white LED strip (Home Accent Lighting Kit, White, PPA International) mounted parallel with the plate walls at a distance of 25 mm.
The assay is very sensitive to inadvertent plate motion and illumination that moves or is poor. Hence, it is important to shield the recording field from ambient light so that the operator's movement does not cast a moving shadow on the field of view.
One of the major stumbling blocks in identifying candidate drugs for the treatment of lymphatic filariasis and river blindness is the lack of a high throughput screening system for these large worms. The filarid nematodes are long and threadlike and cannot be easily assayed in a 96-well format. We therefore developed an automated imaging system in which
The WormAssay is a visual imaging system that utilizes a novel software program to capture video recordings to assay the effect of compounds on macroparasites. To test the robustness of the software program, we assayed
We are currently using the visual imaging system to screen approximately 400 adult
Rather than using laborious and subjective methods of analyzing plates (manual examination of individual wells and plates with a dissecting scope and scoring worm movements relative to control worms), the WormAssay quantified each worm's movement simultaneously on the entire plate, with each plate taking approximately 30 seconds to 1 minute to read. Given the short read times, researchers can increase the number of replicates per compound, thus increasing the accuracy of the assay. Currently, the system requires an individual to place the plate into the visual imaging box but this system is amenable for use with a robotic arm, removing and replacing plates to and from a plate hotel. The software application also includes bar code reading capabilities and can easily be exported to spreadsheets for data analysis.
WormAssay is a unique high-throughput screening motility assay that performs a parallel analysis on each well of entire plates simultaneously, but is independent of specific plate geometry and parasite morphology. The application supports 6-, 12-, 24-, 48- and 96-well plates. WormAssay does not track specific organismal characteristics so it can assay the motility of a large range of macroscopic organisms that can be cultured in a microtiter plate, but is capable of tracking very small or refined movements. The assay requires commodity computer equipment and is compatible with a variety of HD 1080p (or greater resolution) cameras and video capture interfaces. This low-cost and simple-to-use system can also be applied to other target organisms as well. Movements of other macroparasites, including adult schistosome worms were also assessed (see
In summary, the WormAssay offers several advantages: 1) it is inexpensive with costs of the video camera, LED lights and camera totaling less than $3,000 USD and the software is freely available, 2) it is easy to use, i.e. the plate can be quickly placed into the box housing the video camera and removed, 3) video recordings are saved onto the computer along with the data and can be reanalyzed at a later time, 4) entire plates with 6-, 12-, 24-, 48- and 96-wells can be assayed simultaneously, 5) the phenotype (worm movement) is quantified and stored as CSV files and 6) can be more generally applied to the study of macroparasites or other macroscopic organisms.
(MP4)
(MP4)
(MP4)
(MP4)
(MP4)
(MP4)