Optical tracking and laser-induced mortality of insects during flight

In-flight dosing system

The system used in this work is detailed in Fig. 1. It can be broken down into three primary modules: (1) a “coarse tracking” system that uses a pair of stereoscopic cameras to identify the three dimensional location of a target subject, which is passed on to (2) the “fine tracking” system that uses a single higher speed camera and a fast scanning mirror (FSM) to keep the target in the middle of the field of view (FOV) of the camera using a proportional-integral-derivative (PID) control loop, and (3) the laser dosing system that fires the laser pulse, which is co-aligned with the fine tracking system to ensure the laser pulse is accurately applied to the subject even while it is moving. For both the coarse and fine tracking systems, subjects are identified by the size of their silhouettes generated from near infrared LED back-illumination or reflection. As demonstrated in Fig. 1b, the subject cages were 20 cm cubes constructed from clear acrylic, but with the side facing the tracking and dosing systems made of borosilicate glass, placed two meters away from the FSM. A high-speed video camera (Vision Research Phantom) was set up to record a small subset of experiments at 2000 frames per second as well. For all experiments, each cage contained only a single subject to be illuminated, or “dosed,” with the laser, and conditions were set such that a subject could be dosed only when it was at least 2.5 to 5 cm away from any wall of the cage to ensure the subject was flying normally, rather than taking wing or alighting.

Prior to commencing the laser dosing experiments, a number of parameters were defined to analyze the results, as detailed in Table 1 and Fig. 2. The key outcome, in line with our previous study¹⁵ and with WHO guidelines on insecticide treatments¹⁷, was whether the subjects were alive or disabled (i.e. dead or moribund) 24 h after the treatment. To characterize how well the subjects were tracked during the laser pulse, the tracking error was defined as how far the insect’s centroid was from the center of the fine tracking camera’s FOV. Other parameters defined in Table 1 relate to the “occlusion factor,” or how much of the laser beam’s energy (assuming a Gaussian profile) overlapped with the target’s outline, as demonstrated in Fig. 2. From the coarse tracking system’s output of xyz position of the target, we could also define velocity, speed, and both linear and angular acceleration of the subjects before, during, and after the laser pulse.

Initial results with 532 nm laser

Initial experiments were conducted with A. stephensi subjects and a 532 nm laser (Verdi, Coherent) using parameters for power (3 W), pulse duration (25 ms), and laser spot diameter (2.5 mm) that were defined in our previous work¹⁵ as an optimal combination of potential cost and mortality performance. As seen in Supplemental Video 1, the system could be quite effective in disabling a flying mosquito with these parameters. Of note from the video, along with flight tracking data for numerous trials (not shown), the 25 ms pulse duration appeared to be short enough that the mosquitoes did not perceptibly alter their flight pattern during the pulse to throw off the tracking algorithm. From Fig. 3a,b, though, the initial system did not have consistent enough tracking performance, such that survival of the subjects was almost entirely dictated by how well the fine tracking system kept the target near the center of its FOV. From this initial experiment on a sample of 80 subjects, we set limits for mean and maximum tracking error during the laser pulse as 2 and 3 pixels, respectively, where each pixel represents ~ 250 μm. Figure 3c,d demonstrates that the system was able to achieve and maintain this performance after modifying the PID loop parameters, and that there was no longer any association between tracking performance and mortality at these conditions. These criteria were evaluated and assured for all mortality data reported in this manuscript (i.e. a Kruskal–Wallis test indicated no significant differences in tracking errors among subjects that survived or were disabled by the laser pulses). Further, Supplemental Fig. S1 shows that flight behavior, in particular speed and linear acceleration, did correlate with tracking accuracy, but that the system was able to meet the noted tracking requirements even at the extremes of flight behavior that could be observed in this study given the restricted flight volumes (though the typical values seen here largely align with available data for other Anopheles mosquitoes^18,19).

Mortality results were analyzed as in Fig. 4. Each data point represents the mortality outcomes (i.e. percentage of targets that were dead or moribund at 24 h, assuming acceptable control mortality < 5%) of 40 subjects, dosed one at a time with one subject per cage, at a given fluence (laser pulse energy per unit area). Logistic regression curves were then fit to the data to determine LD90 (lethal dose 90, or dose where 90% mortality is expected) for a given set of laser conditions. Figure 4 shows the resulting dose–response curve for the parameters described above—532 nm, 3 W, 25 ms pulse duration, and 2.5 mm spot diameter. For comparison, the dose–response curve using these same parameters for anesthetized mosquitoes in our previous work¹⁵ is shown as well. Although confidence intervals for the anesthetized fit are not shown for the sake of visual clarity, it is apparent that there were no substantial differences in the dose–response curves for the anesthetized and the in-flight dosing experiments, which utilized the same stock of A. stephensi from the same source, and otherwise cared for identically.

Effects of larger spot size and shorter pulse duration

As shown in Table 2 and the left-hand side of Fig. 5, additional experiments explored the impacts of varying other parameters with the same 532 nm laser, in particular examining whether shorter pulse durations or larger spot sizes would result in net reductions of the LD90 fluence. Combinations of both using a higher power with shorter pulse durations, or a higher power but larger spot diameter (to 4 mm) led to slight increases in the LD90. Of note, the 6 mm spot diameter offered limited data points due to the maximum power available from the laser. Figure 5 also shows two curves using a 3.5 W blue diode laser (445 nm, Nichia NDB7K75), which would be a more cost-effective implementation source compared with the lab-grade 532 nm Coherent laser. The two sources were expected to provide similar laser-mosquito interaction regimes, and indeed, Fig. 5 shows that the dose–response curves for both sources are nearly identical when implemented with 3 W, 2.5 mm spot diameters, and 25 ms pulse durations. Because of the limited blue diode power, the spot diameter was not increased, but one experiment was completed with a 1.4 mm spot diameter and 25 ms pulse duration; this dose–response curve looked markedly different from the other visible wavelength curves in Fig. 5, with an LD90 > 2 × that of the other experiments at 445 or 532 nm. Note that this greater LD90 fluence value for the smaller spot still represented a lower pulse energy than required for the larger spots given the spot size differential.

Dosing with near infrared wavelengths

Two near infrared (NIR) laser sources were examined as well—a custom constructed 1,064 nm (1 μm) fiber laser with a maximum output of 30 W and a commercial 1,570 nm (1.5 μm) fiber laser (IPG Photonics) with a maximum deliverable output of 11 W. The right-hand side of Fig. 5 shows the dose–response curves for these laser sources in addition to those using visible wavelengths. All of the infrared experiments resulted in significantly higher LD90 fluence levels compared with the visible lasers, and given the log scale of Fig. 5, varying experimental conditions with the 1 μm source led to comparatively larger changes in LD90 compared with the visible sources. As with the blue diode, the small spot (1.5 mm) had the highest LD90 fluence, but unlike the visible lasers, the larger spot (4 mm) offered a lower LD90 fluence compared with the baseline 2.5 mm. For the 1.5 μm source, there was insufficient mortality at the maximum power available from this source with a 2.5 mm spot and 25 ms pulse duration, so Fig. 5 shows a curve using slightly longer pulse durations to make up the additional fluence required. From the available data, it appears that the LD90 fluence levels for the 1 μm and 1.5 μm sources are at least at reasonably comparable levels.

Effects of longer pulse durations with lower power

We then further explored the effects of increasing the pulse duration to determine whether this could relax the optical power required from the laser source, given that laser cost is correlated with output power. Figure 6a,b show the results for the 532 nm and 1,064 nm sources, respectively, with spot diameters of 2.5 mm and pulse duration set to 25, 50, or 100 ms. In all cases, the optical power was adjusted according to the pulse duration to supply a constant pulse energy, and therefore fluence at approximately the LD75 level determined from previous experiments, which was chosen to ensure a low likelihood of any data point being 100% mortality (i.e., a saturated signal). As evidenced from Fig. 6a and Table 3, at 532 nm there was a significant drop off in mortality at the longer pulse durations, although the effect is smaller going from 50 to 100 ms compared with the initial drop from 25 to 50 ms. Similar, but smaller magnitude effects were seen with the 1 μm source in Fig. 6b and Table 3. Supplemental Fig. S2 shows that the tracking accuracy for these experiments somewhat mirrors the mortality performance, with a notable decrement from 25 to 50 ms but no significant change from 50 to 100 ms.

Comparing dosing performance across other insects

Using the same system and procedures, experiments were also run on SWD and ACP subjects as a way of cursorily exploring laser-insect interaction for species of interest besides A. stephensi. For SWD, a full dose–response curve was created for the 1 μm laser with typical exposure settings (2.5 mm spot diameter and 25 ms pulse duration). The LD90 fluence from this work was 8.6 J/cm², compared with 12.9 J/cm² for A. stephensi under the same conditions. Tracking performance on SWD subjects was comparable to that for A. stephensi, given that flight parameters such as speed and acceleration were comparable as well. With ACP subjects, inducing them to fly in the cages was very difficult. As such, we could not process a sufficient number of subjects to create a proper dose–response curve. Based on the limited data available, and as demonstrated in Supplemental Video 2, the system as set for the mosquito LD90 condition with the 1 μm laser, 2.5 mm spot diameter and 25 ms pulse duration was effective in tracking and disabling the ACP subjects, despite their very different flying behavior (greater speeds and accelerations, limited durations) relative to the other test species.

Longer range dosing system demonstration

As a final proof of principle test, a long-range version of the system from Fig. 1 was configured. This system worked at a distance of 30 m instead of 2 m, facilitated primarily by modifying the optics used for the coarse and fine tracking systems. Given the reduced flexibility of this system, it was only used with the 1 μm laser with a 2.5 mm spot diameter and 25 ms pulse duration. Rather than acquiring new dose–response curves, which was impractical given the setup’s location in a building ~ 30 km away from the insect-housing chamber, we verified the efficacy of the system using the LD90 conditions established by short-range dosing. Supplemental Video 3 shows this system dosing an A. stephensi subject from the same stock as used for short-range testing, and Supplemental Video 4 shows the same for a wild-sourced Culex pipiens mosquito obtained in a local pond. Note that the wild C. pipiens was substantially larger in size than the lab-reared A. stephensi. In all of the limited number of tests of this system for both species (10 A. stephensi and 5 C. pipiens), the mortality rate was 100%.

Source: Ecology - nature.com