Dataset Information

Deep learning improves taphonomic resolution: high accuracy in differentiating tooth marks made by lions and jaguars

ABSTRACT: Taphonomists have long struggled with identifying carnivore agency in bone accumulation and modification. Now that several taphonomic techniques allow identifying carnivore modification of bones, a next step involves determining carnivore type. This is of utmost importance to determine which carnivores were preying on and competing with hominins and what types of interaction existed among them during prehistory. Computer vision techniques using deep architectures of convolutional neural networks (CNN) have enabled significantly higher resolution in the identification of bone surface modifications (BSM) than previous methods. Here, we apply these techniques to test the hypothesis that different carnivores create specific BSM that can enable their identification. To make differentiation more challenging, we selected two types of carnivores (lions and jaguars) that belong to the same mammal family and have similar dental morphology. We hypothesize that if two similar carnivores can be identified by the BSM they imprint on bones, then two more distinctive carnivores (e.g. hyenids and felids) should be more easily distinguished. The CNN method used here shows that tooth scores from both types of felids can be successfully classified with an accuracy greater than 82%. The first hypothesis was successfully tested. The next step will be to differentiate diverse carnivore types involving a wider range of carnivore-made BSM. The present study demonstrates that resolution increases when combining two different disciplines (taphonomy and artificial intelligence computing) in order to test new hypotheses that could not be addressed with traditional taphonomic methods.

SUBMITTER: Jimenez-Garcia B

PROVIDER: S-EPMC7423429 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Camera trapping is the standard field method of monitoring cryptic, low-density mammal populations. Typically, researchers run camera surveys for 60 to 90 days and estimate density using closed population spatially explicit capture-recapture (SCR) models. The SCR models estimate density, capture probability (g0), and a scale parameter (?) that reflects ranging behaviour. We used a year of camera data from 20 camera stations to estimate the density of male jaguars (Panthera onca) in the Cockscomb Basin Wildlife Sanctuary in Belize, using closed population SCR models. We subsampled the dataset into 276 90-day sessions and 186 180-day sessions. Estimated density fluctuated from 0.51 to 5.30 male jaguars / 100 km2 between the 90-day sessions, with comparatively robust and precise estimates for the 180-day sessions (0.73 to 3.75 male jaguars / 100 km2). We explain the variation in density estimates from the 90-day sessions in terms of temporal variation in social behaviour, specifically male competition and mating events during the three-month wet season. Density estimates from the 90-day sessions varied with ?, but not with the number of individuals detected, suggesting that variation in density was almost fully attributable to changes in estimated ranging behaviour. We found that the models overestimated ? when compared to the mean ranging distance derived from GPS tracking data from two collared individuals in the camera grid. Overestimation of ? when compared to GPS collar data was more pronounced for the 180-day sessions than the 90-day sessions. We conclude that one-off ('snap-shot') short-term, small-scale camera trap surveys do not sufficiently sample wide-ranging large carnivores. When using SCR models to estimate the density from these data, we caution against the use of poor sampling designs and/or misinterpretation of scope of inference. Although the density estimates from one-off, short-term, small-scale camera trap surveys may be statistically accurate within each short-term sampling period, the variation between density estimates from multiple sessions throughout the year illustrate that the estimates obtained should be carefully interpreted and extrapolated, because different factors, such as temporal stochasticity in behaviour of a few individuals, may have strong repercussions on density estimates. Because of temporal variation in behaviour, reliable density estimates will require larger samples of individuals and spatial recaptures than those presented in this study (mean +/- SD = 14.2 +/- 1.2 individuals, 37.7 +/- 4.7 spatial recaptures, N = 276 sessions), which are representative of, or higher than published sample sizes. To satisfy the need for larger samples, camera surveys will need to be more expansive with a higher density of stations. In the absence of this, we advocate longer sampling periods and subsampling through time as a means of understanding and describing stability or variation between density estimates.

Project description:The accuracy of motion prediction, utilized to overcome the system latency of motion management radiotherapy systems, is hampered by irregularities present in the patients' respiratory pattern. Audiovisual (AV) biofeedback has been shown to reduce respiratory irregularities. The aim of this study was to test the hypothesis that AV biofeedback improves the accuracy of motion prediction.An AV biofeedback system combined with real-time respiratory data acquisition and MR images were implemented in this project. One-dimensional respiratory data from (1) the abdominal wall (30 Hz) and (2) the thoracic diaphragm (5 Hz) were obtained from 15 healthy human subjects across 30 studies. The subjects were required to breathe with and without the guidance of AV biofeedback during each study. The obtained respiratory signals were then implemented in a kernel density estimation prediction algorithm. For each of the 30 studies, five different prediction times ranging from 50 to 1400 ms were tested (150 predictions performed). Prediction error was quantified as the root mean square error (RMSE); the RMSE was calculated from the difference between the real and predicted respiratory data. The statistical significance of the prediction results was determined by the Student's t-test.Prediction accuracy was considerably improved by the implementation of AV biofeedback. Of the 150 respiratory predictions performed, prediction accuracy was improved 69% (103/150) of the time for abdominal wall data, and 78% (117/150) of the time for diaphragm data. The average reduction in RMSE due to AV biofeedback over unguided respiration was 26% (p < 0.001) and 29% (p < 0.001) for abdominal wall and diaphragm respiratory motion, respectively.This study was the first to demonstrate that the reduction of respiratory irregularities due to the implementation of AV biofeedback improves prediction accuracy. This would result in increased efficiency of motion management techniques affected by system latencies used in radiotherapy.

Dataset Information

Deep learning improves taphonomic resolution: high accuracy in differentiating tooth marks made by lions and jaguars

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets