The effects of range and echo-phase on range resolution in bottlenose dolphins (Tursiops truncatus) performing a successive comparison task^a)

Determining the distance (range) to objects in the environment is a critical feature of animal biosonar. The primary cue that echolocating animals use to resolve object range is the echo-delay, the time it takes for the emitted acoustic signal to travel to the target and its echo to return to the animal (Simmons and Vernon, 1971; Simmons, 1973; Murchison, 1980; Au, 1993). The delay corresponding to the first echo from a target allows the animal to determine the range to the target, whereas echoes from multiple reflectors within a target and separated by small time scales (i.e., up to hundreds of microseconds) allow the animal to determine spatial features of the target (Simmons , 1990b; Au, 1993). By using large and small time-scale echo-delays, echolocating animals can quickly detect and classify objects within their environment at a performance rate higher than any man-made system (Roitblat , 1995; Moore, 1997; Vishnu , 2022).

Past studies of echo-delay resolution in microchiropteran bats required the bats to treat the task as either a successive or simultaneous comparison task (Simmons and Vernon, 1971; Simmons, 1973; Roverud and Grinnell, 1985; Simmons and Grinnell, 1988; Masters and Jacobs, 1989; Mogdans and Schnitzler, 1990). A successive task requires the animal to echo-inspect one target and store that information in memory and then compare it to a second target. In a simultaneous task, the animal is presented both targets simultaneously and can capitalize on cues created by the reflection of both targets from the same incidental echolocation pulse (Simmons, 1973; Roverud and Grinnell, 1985; Simmons and Grinnell, 1988; Masters and Jacobs, 1989). Simmons (1973) trained the big brown bat (Eptesicus fuscus) on a two-alternative forced choice (2AFC) task in which it inspected a target to the right and then compared it to a target to the left and reported which target was closer. The angular separation between the two targets was 40°, which allowed the bat to ensonify both targets with a single pulse (Simmons, 1973, 2014). Four different absolute ranges (30, 60, 120, and 240 cm) were tested. For the remainder of this publication, absolute range of the targets relative to the echolocator will be referred to as “range” and difference in range between the two targets as “ΔR”. During threshold testing, the range of the farther (S−) target was held constant, while the S+ target was shifted closer to the bat (e.g., at a range of 30 cm, S+ would be shifted between 20 and 29 cm, while S− was always presented at 30 cm). Discrimination thresholds (75% correct) were ∼1.2–1.4 cm (equivalent echo-delays of ∼60–80 μs) and did not vary significantly across the four ranges tested (Simmons, 1973). The experiment was replicated with a phantom echo generator (PEG) at the 30-cm range to present different echo-delays while removing other range parameters (i.e., echo amplitude and spectral parameters). Performance and thresholds for both physical objects and the PEG were similar, suggesting that the bats determined the range of the targets by the arrival time of the corresponding echoes (Simmons, 1973). Additional experiments where a single range was tested with different bat species produced similar discrimination thresholds of ∼1 cm (Roverud and Grinnell, 1985; Surlykke and Miller, 1985; Masters and Jacobs, 1989).

Murchison (1980) replicated the bat experiment conducted by Simmons (1973) with a bottlenose dolphin (Tursiops truncatus) but reached a different conclusion. Rather than finding discrimination thresholds constant at all ranges, thresholds increased with an increase in range. Like Simmons (1973), a 2AFC paradigm was used where the dolphin ensonified two 7.62-cm diameter foam spheres with target strengths (TSs) of −34 dB and reported which sphere was closer in range. The two spheres were arranged with 40° of angular separation relative to the dolphin at ranges of 1, 3, and 7 m. Differing from Simmons (1973), the closer target (S+) remained constant, while the farther target (S−) changed in range (e.g., at a range of 1 m, S+ would always be presented at 1 m, while S− was presented between 1.01 and 1.1 m). Discrimination thresholds (ΔR at 75% correct) were 0.9, 1.5, and 3 cm (12, 20, and 40 μs echo-delay) at ranges of 1, 3, and 7 m, respectively, indicating that as range increased, the resulting ΔR at threshold also increased. Whether the dolphin completed the task using a simultaneous or successive process is unknown. Moore (2008) suggests that dolphins can detect the presence of a spherical target with a TS of –27 dB and when azimuthal angles from the center of the dolphin's beam are greater than ±20°. However, with the lower TS, –34 dB, of the targets used in Murchison (1980) and the narrow beam width of high-frequency content in the dolphin's echolocation beam, it is questionable whether the dolphin could have extracted useable information from both targets with a single click (Au , 1986; Finneran , 2014; Finneran , 2016). Therefore, whether differences in performance as a function of range in the bottlenose dolphin versus the bat represent a true difference in sensory system capabilities or simply differences in experimental design is unknown.

The current experiment tested dolphin range discrimination thresholds in a successive comparison task using a two-channel PEG system. Testing was conducted in a 2AFC paradigm at seven different ranges from 1.75 to 20 m. Additionally, at a range of 7 m, threshold testing was conducted using both constant echo-phase and a random phase shift (±180°) that was assigned to each echo. Previous research suggests that big brown bats and dolphins appear sensitive to changes in echo-phase during jitter delay experiments (Simmons, 1979; Menne , 1989; Simmons , 1990a; Moss and Simmons, 1993; Finneran , 2019; Finneran , 2020; Finneran , 2023). However, studies with pale spear-nosed bats (Phyllostomus discolor) where echo-phase was changed by manipulating the phantom target impulse response duration, rather than delay, suggest that these bats could not use the complete echo-phase spectrum but only the portion that encodes echo-delay (Schörnich and Wiegrebe, 2008). Whether changes in echo-phase affect the dolphin's ability to discriminate ranges between targets is unknown. The random phase shift had the effect of altering the fine structure of each echo waveform to shift the positions of the waveform peaks and valleys from echo to echo—without changing the waveform envelope. By removing consistent fine echo structure, only the envelope of the echo waveform could be used to determine range. These experiments determined discrimination thresholds as a function of range and echo-phase that had not been previously identified.

Two bottlenose dolphins participated in the study: Eclipse (ECL) (male, 7 years) and Lark (LRK) (female, 17 years). Upper-frequency limits (UFLs) for their hearing, defined as the frequency at which electrophysiological auditory steady-state response thresholds reached a sound pressure level (SPL) of 120 dB re 1 μPa, were ∼56 kHz for ECL and ∼136 kHz for LRK. ECL was considered to have high-frequency hearing loss, and LRK was considered to have full-bandwidth hearing, which was defined as a hearing range with UFL ≥ 120 kHz (Johnson, 1966; Houser and Finneran, 2006; Strahan , 2020).

Tests were conducted in a 9 × 9 m floating netted enclosure at the U.S. Navy Marine Mammal Program in San Diego Bay, CA, between September 2021 and July 2022. During each trial, the dolphin positioned itself in an underwater polyvinyl chloride (PVC) “hoop station” located at ∼1 m depth and supported by a single vertical post. Response paddles were located ∼0.3 m to the dolphin's left and right. The hoop station was oriented so that the dolphin faced San Diego Bay through a netted frame (Fig. 1). Beyond the frame, at 1.2 m distance and ±20° azimuthal angle from the center-point of the hoop station, were two pairs of piezoelectric transducers (TC4013, Reson Inc., Slangerup, Denmark). Each transducer pair operated as one of two independent channels (left/right) in the PEG system. One transducer in each pair served as the dolphin's biosonar click receiver and the other as the echo projector for the associated channel. An additional piezoelectric transducer was embedded in a silicone suction cup and placed in the center of the dolphin's melon, approximately 3 cm above its rostrum. The hydrophone on the melon was used to estimate the time of click emission (see below). The nearest underwater structures within ±20° of the dolphin's main biosonar transmission axis while in the hoop station were ∼500 m distant. The mean water depth was ∼10 m. Ambient noise consisted of snapping shrimp, vessel traffic, and other dolphins. Median ambient noise pressure spectral density levels were approximately 69 dB re 1 μPa²/Hz at 10 kHz and decreased linearly with the logarithm of frequency to 52 dB re 1 μPa²/Hz at 150 kHz.

The dolphin's task was to position itself in the hoop and produce echolocation clicks toward the left or right click receiver while listening to returning phantom echoes from the corresponding projector. After ensonifying both sides, the dolphin was required to leave the hoop and touch the response paddle corresponding to the side (left or right) that produced the phantom echoes with the shorter echo-delay (S+). Each dolphin participated in one session per day, and sessions were run daily, Monday–Friday. Sessions lasted ∼25 min and consisted of ∼60 individual trials. Each trial began with the PEG off and the dolphin directed to position itself in the hoop. Once positioned, the PEG was turned on for 4 s, with each PEG channel simulating a target at a different range. The dolphin could leave the hoop anytime during the 4-s trial to make its choice. Trials where the dolphin left the hoop prior to triggering echoes from both the left and right channels and trials where the dolphin did not touch a paddle were not included for analysis; these represented 1.3% of the total trials.

Thresholds were measured at seven different simulated ranges with no phase shift and at the 7-m range with each echo given a random phase shift. Data collection for each range took ∼3–5 weeks and occurred in the following order: 10, 5, 7, 3, 1.75, 14, and 20 m, followed again by 7 m with random phase shifts. To ensure the dolphins were comparing the range of both simulated targets versus using a template of the S+ range stored in memory, the S+ and S− ranges roved around the ranges listed above from trial to trial. The specific range of the S+ on a trial-to-trial basis was selected from a distribution in which the mean was defined by the range being tested and with the standard deviation set to 15 cm. Random selection of the S+ range was limited to a normal distribution truncated at ±3 standard deviations. At 1.75-m range, roving was limited to a minimum range of 1.4 m due to the physical distance to the transducers and the minimum time required by the PEG to generate an echo. This distribution still allowed for distinct separation of distributions between each range tested.

Seven or eight ΔRs were tested each session depending on range. Each session began with warm-up trials, where each unique value of ΔR was presented to the dolphin in descending order across trials. After the warm-up, ΔR was randomized from trial to trial, while ensuring an equal number of trials on the left and right for each ΔR. At least 36 trials were conducted at each range and ΔR combination, except for ECL at 20 m (28 trials at ΔR = 22 and 65 cm and 32 trials at ΔR = 45 cm), where he demonstrated a decrease in motivation and willingness to participate when the range was set to 20 m.

Relative echo level [REL, the echo sound exposure level (SEL, in dB re 1 μPa²s) at the listening position relative to the click SEL at the click receiver] was held constant as the range varied. REL was set to 15 dB above ECL's detection threshold, which was also the highest possible REL for the hardware. As REL was independent of simulated range, apparent TS increased with increasing range (Table I). Due to his high-frequency hearing loss, ECL's effective TSs were estimated by applying a low-pass filter to the echo at his UFL of 56 kHz. REL was set to 35 dB above LRK's detection threshold, resulting in a REL 14 dB less than ECL's REL. These adjustments were applied between subjects to produce similar TSs for each dolphin (Table I).

The PEG was implemented using an NI PXIe-7856R device containing a Kintex-7 160 T FPGA. Clicks emitted by the dolphin were captured by the left, right, and melon click receivers, amplified, filtered (5–200 kHz, VP-1000, and 3C module, Krohn-Hite Corp., Brockton, MA), and then digitized by the PXIe-7856R with 1-MHz sampling rate and 16-bit resolution. If the digitized hydrophone signal exceeded an amplitude threshold, click waveforms were extracted from the left and right channels, and times of arrival were obtained from all three channels. Three criteria were used to determine whether the digitized click signal from the left or right channel “triggered” the PEG and resulted in echo generation from that side: (1) the digitized hydrophone signal on that channel exceeded the amplitude threshold; (2) the click peak-to-peak (p-p) amplitude on that channel (left or right) was greater than the amplitude of the same click on the other channel; and (3) the time-of-arrival difference (TOAD) between the click received by the left/right hydrophone and the same click received by the melon hydrophone was less than the nominal acoustic travel time for that simulated target range. Criterion (2) ensured that echoes were only generated from a single channel, the side the dolphin's beam was primarily directed toward. Criterion (3) ensured that the dolphin remained in the hoop during the echo inspection. Click waveforms that triggered the PEG were convolved with a target impulse response function to create the echo waveform. The echo waveform was then scaled in amplitude, delayed, and converted to analog (PXIe-7856R, 1 MHz, 16 bit). To minimize effects of dolphin head movement within the hoop, the delay of each echo was corrected using the TOAD between the click received by the left/right hydrophone and the same click received by the melon hydrophone (an indication of the dolphin's instantaneous position relative to the click receiver). The analog echo waveform was then filtered (5–200 kHz, 3C module), amplified (M7600, Krohn-Hite Corp.), and used to drive the echo transmitter. The hydrophone signals and echo waveforms were also digitized at 2 MHz and 16-bit resolution by an NI PXIe-6368 multifunction data acquisition device (National Instruments, Austin, TX) and stored for later analysis (Fig. 2).

Convolution was performed in the frequency domain using a fractional delay technique (Välimäki and Laakso, 2000) to obtain echo-delay resolutions below the 1-μs sampling interval. Target transfer functions included only phase and delay elements. Larger-scale echo-delay, in integral multiples of 1 μs, was achieved by changing the position of the echo waveform in the D/A converter output buffer. In this fashion, echo-delays up to ∼200 ms could be achieved with resolution < 0.001 μs. In practice, echo-delay resolution was limited by the inherent “jitter” in the system, primarily arising from motion of the transducers relative to the hoop (e.g., caused by water motion). The inherent jitter, estimated by repeatedly triggering the PEG using a representative electronic click waveform and measuring the delay of the resulting acoustic echoes in the center of the hoop, was ±0.4 μs.

Operation of the PEG was verified before each session by replacing the dolphin click signal input to the PEG A/D converter with a representative recording of an on-axis dolphin click. Calibrations were performed prior to each session by broadcasting the analog echo waveforms from the left echo projector and recording the acoustic echoes (without the dolphin present) with the right click receiver; this method was then reversed (right projector to left click receiver). Calibrations over the course of the study varied by ±1 dB.

Statistical analysis was conducted in R (R Core Team, 2019). The quickpsy package was used to build the psychometric functions using the cumulative normal distribution function to fit proportion correct versus ΔR for each range (Linares and López-Moliner, 2016). Model parameters were estimated using maximum likelihood. Threshold was calculated at the 75% correct rate from the psychometric function, and the 95% confidence intervals were calculated by bootstrapping the data with 1000 iterations. A linear mixed model was built with absolute range and subject as main effects; however, subject was non-significant (p = 0.609) and, therefore, removed from the model. Data for the subjects were subsequently combined.

Custom software was used to calculate click acoustic parameters including inter-click interval (ICI), p-p SPL, center frequency, and (centralized) root mean square (rms) bandwidth. To restrict analysis to clicks near the main transmit axis, only clicks within ±3 dB of the maximum click p-p SPL for each individual trial were included in the analysis—these were considered on-axis clicks. Representative echoes were measured by inputting the mean on-axis click waveform of each subject into the PEG and then broadcasting the resulting echo and measuring the acoustic pressure at the center of the hoop. Measurements were repeated 1024 times, and the acoustic pressure was synchronously averaged to calculate the echo waveform. Effects of ECL's high-frequency hearing loss on received echo levels were estimated by low-pass filtering the echo at 56 kHz (8th order Butterworth, zero phase) to obtain the “effective” echo waveform and spectrum.

Preliminary analysis displayed differences in LRK and ECL's click production, and therefore clicks were analyzed independently for each dolphin. This resulted in 1.2 × 10⁴ clicks analyzed for ECL and 1.5 × 10⁴ clicks analyzed for LRK. The mean click waveform and spectra for each dolphin showed characteristics of a normal dolphin click [Fig. 3(a)]. Projected echoes [Fig. 3(b)] were longer in duration and contained more high-frequency content than the dolphin's clicks due to the characteristics of the projector transmitting response. Echo amplitudes for LRK were 14 dB lower than those for ECL; however, low-pass filtering using ECL's upper hearing limit resulted in an effective echo with lower amplitude [see Fig. 3(b)].

Distributions of ICIs and p-p SPLs (dB re 1 μPa) for each dolphin were similar when inspecting the left and right PEGs. Therefore, clicks from the left and right PEGs were combined for analysis of ICIs and p-p SPLs (Fig. 4). ICI distributions were broad, especially at the longer ranges, due (in part) to roving the ranges of S+/S− and the multiple values of ΔR. ECL's ICIs varied from ∼8 to 50 ms and generally increased with target range, except for the 5-, 7-, and 10-m ranges, where ICIs were similar. LRK's ICIs varied from ∼3 to 40 ms and increased systematically with target range. ECL's click p-p SPLs varied from ∼195 to 215 dB re 1 μPa and did not change systematically with a change in a range. LRK's click p-p SPLs varied from 190 to 215 dB re 1 μPa and tended to increase with target range.

Distributions of center frequency and rms bandwidth for each dolphin differed between the right and left click receivers but remained consistent for different ranges. Therefore, clicks from the left and right PEGs were analyzed separately for the left and right receivers, but absolute ranges were combined for each dolphin (Fig. 5). ECL and LRK both had fewer clicks analyzed on the right compared to the left because there were fewer on-axis clicks on the right. ECL's center frequencies were between 50 and 100 kHz for the left PEG and between 55 and 90 kHz for the right PEG, while LRK's center frequencies were between 75 and 115 kHz for the left PEG and between 65 and 105 kHz on the right PEG. ECL's rms bandwidths were between 25 and 40 kHz for the left PEG and between 20 and 35 kHz on the right PEG, while LRK's rms bandwidths were between 25 and 40 kHz for the left and right PEGs (Fig. 5).

Figure 6 shows the performance of ECL and LRK at each range for the normal phase condition. Due to the lack of significant differences between subjects, data were combined to calculate psychometric functions. Thresholds were calculated by interpolating along the psychometric function to find the ΔR where the dolphins were 75% correct. Figure 7 shows the resulting thresholds as a function of range, with error bars representing the 95% bootstrapped confidence intervals. Data were fit well with a piecewise linear function, using the segmented package in R (Muggeo, 2003). The resulting slopes were 0.67 and 7.15 cm/m with a breakpoint of 11 m (R² = 0.99).

Figure 8 compares performance at 7-m range between the normal phase condition (i.e., no echo-phase shifts) and when each echo was given a random phase shift. Thresholds for the normal and random phase shift conditions were 8.27 and 8.73 cm, respectively. A linear mixed model was built to determine whether the phase shift had a main effect, and the model determined no significant differences between the normal phase and random phase shift conditions (p = 0.171).

The present experiment determined discrimination thresholds as a function of range, at farther ranges than previous dolphin and bat experiments, using phantom echoes rather than physical targets. Therefore, echo-delay could be precisely controlled and manipulated independently from echo amplitude. In the current study, range discrimination thresholds were 4.1, 6.5, and 8.3 cm at 1.75-, 3-, and 7-m range, respectively, equating to echo-delay differences of 55, 87, and 110 μs. Thresholds were consistently higher than those reported by Murchison (1980): 0.9, 1.5, and 3 cm at 1-, 3-, and 7-m range (echo-delay differences of 12, 20, and 40 μs), respectively. Present results are also much larger than echo-delay thresholds obtained using a jittered-echo paradigm (Finneran , 2023). This suggests that a different mechanism is used for jitter detection than for successive ranging and that higher thresholds are not solely a result of the animal's head movement (Altes, 1989).

There could be several reasons for the high thresholds compared to Murchison (1980). Murchison (1980) used physical targets, and therefore the TS was constant with range, but echo-delay and relative echo level were coupled. In contrast, in the present study, relative echo level was held constant to decouple echo-delay and echo level as much as possible. This results in the apparent TS decreasing as range decreases, resulting in much lower TSs in the present study than in Murchison (1980). The TS of the foam spheres in the previous study was −34 dB, whereas the simulated TS in the current study for ranges 1.75–7 m for LRK was –81 to –57 dB, and the effective TS for ECL was −87 to –63 dB. Although these TSs resulted in echoes above each dolphin's echo detection threshold, the dolphin in the previous study could have benefitted from the higher echo signal-to-noise ratio (SNR) resulting from the higher TS. Additional research would need to be conducted to determine how differences in SNR affect range resolution in dolphins. The foam spheres used in Murchison (1980) also would have reflected a complex echo, whereas in the current experiment, the echoes were a replica of the dolphin's click representing a point target. The higher TS and simultaneous presentation of targets paired with the complex echo may have enabled the dolphin to utilize spectral interference patterns resulting from near-simultaneous reception of reflections of the same click from both targets as he scanned from one target to the other. Additional research is needed to determine whether the dolphin's performance would improve during a simultaneous process of range discrimination versus the successive process that was investigated in the current study.

Murchison (1980) concluded that the range discrimination thresholds measured followed the Weber–Fechner function, defined as ΔR/R = K, where R is the absolute range and K is a constant. In the current experiment, the data fit this function up until the 10-m absolute range. Previous research suggests that different mechanisms are used for target identification at ranges greater than 14 m (Finneran , 2013). In the current experiment, performance degrades rapidly at ranges of 14 m and greater. Given the limited data above 10 m, it is not clear how thresholds continue to change as range continues to increase and whether there is an upper limit beyond which they cannot discriminate between ranges. It has been reported at ranges greater than 75 m that dolphins begin using click “packets,” where the dolphin projects a burst of clicks and waits for the corresponding collection of echoes to return before sending out the next burst or set of packets (Ivanov and Popov, 1978; Ivanov, 2004; Finneran, 2013; Ladegaard , 2019). Thus, measuring range resolution at large ranges could be affected by an overall change in echolocation strategy beyond a certain target range.

Of note are the apparent differences among the current study, Murchison (1980), and the experiment conducted in bats by Simmons (1973). Simmons (1973) reported discrimination thresholds for ranges from 30 to 240 cm (echo-delays of 1.8–14.1 ms) to be ∼1.2–1.4 cm (∼60–80 μs) and reported little change in threshold with a change in range. However, Simmons (1973) did not correct for head movement. Simmons and Grinnell (1988) suggested that results from the 1973 experiment could be better characterized as measurement of the bat's head movement versus the ability to resolve range. If this were the case, the error from head movement would be similar as range increased, leading to similar thresholds at each range. In the current study, head movement was accounted for with a melon hydrophone, while in Murchison (1980), head movement was controlled by the dolphin placing its rostrum in a “chin cup” that allowed its head to scan right to left but controlled the range to each target. The restriction of head movement could therefore account for the threshold differences as a function of range among the bat and dolphin studies.

In the current study, there were no significant differences between the psychometric functions for normal and random phase conditions at 7-m range. This result is perhaps not surprising given the relatively large threshold at 7 m (i.e., larger than the echo envelope). Since randomizing the phase disrupts the fine structure, the random phase data further highlight that the echo envelope is utilized for determining range in a successive task.

A unique opportunity of the present study was to compare click emissions from LRK, with full hearing bandwidth, and ECL, who had high-frequency hearing loss with an upper-frequency hearing limit of 56 kHz. Although there were no significant differences between their behavioral data, there were apparent differences between click emissions. The average center frequency for ECL's click was ∼20 kHz less than LRK's (although clicks for ECL still contained frequencies above the subject's hearing range). There was a systematic increase in ICI with an increase in range for both subjects. However, at the closest range of 1.75 m, LRK's ICIs were ∼4 ms, whereas ECL's are much higher at ∼14 ms. LRK also decreased click p-p SPL at 1.75 m, and the p-p SPL continued to increase with range, although relative amplitude remained constant for all ranges. Finneran (2013) reports similar results during an echolocation change detection task where REL was held constant at different simulated ranges. It is uncertain whether this is a natural response of the dolphin to increase click amplitude as range increases (Au and Benoit-Bird, 2003; Jensen , 2009) even in cases where REL is artificially held constant (Finneran , 2013). Interestingly, the p-p SPL of ECL's click varied slightly at the different ranges but did not change steadily with range. There is potential that this subject's high-frequency hearing loss was a contributor to this difference, as he could not decrease the p-p SPL of his clicks to the same level as LRK and still audibly hear the return echo. Additionally, ECL (age 7) is a less experienced dolphin than LRK (age 17), and their strategies to complete the task could have been affected by their prior experience.

As target range increases, range discrimination thresholds in dolphins increase during a successive range discrimination task. In contrast to “jittered” delay resolution tasks, range discrimination in a successive comparison does not depend on echo fine structure, but rather on the echo envelope.

The authors thank H. Bateman, K. Bucknam, and T. Wu, who contributed greatly to the daily training and data collection for both subjects; R. Dear, D. Werneth, the marine mammal trainers, and the animal care and training interns with the Navy Marine Mammal Program for their daily support and coordination; and members of the Gentner Lab for their feedback throughout the study. Financial support was provided by the Office of Naval Research Code 322. This is National Marine Mammal Foundation Contribution #380 to peer-reviewed scientific literature.

The authors have no conflicts of interest to disclose.

The study followed a protocol approved by the Institutional Animal Care and Use Committee at the Naval Information Warfare Center Pacific and the Navy Bureau of Medicine and Surgery and followed all applicable U.S. Department of Defense guidelines.

The data that support the findings of this study are available from the corresponding author upon reasonable request.