Laser ablation mass spectrometry blast through detection in R.

RATIONALE: Organisms that grow a hard carbonate shell or skeleton, such as foraminifera, corals or molluscs, incorporate trace elements into their shell during growth that reflect the environmental change and biological activity they experienced during life. These geochemical signals locked within the carbonate are archives used in proxy reconstructions to study past environments and climates, to decipher taxonomy of cryptic species and to resolve evolutionary responses to climatic changes. METHODS: Here, we use laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) as a time-resolved acquisition to quantify the elemental composition of carbonate shells and skeletons. We present the LABLASTER (Laser Ablation BLASt Through Endpoint in R) package, which imports a single time-resolved LA-ICP-MS analysis, then detects when the laser has ablated through the carbonate as a function of change in signal over time and outputs key summary statistics. We provide two examples within the package: a fossil planktic foraminifer and a tropical coral skeleton. RESULTS: We present the first R package that automates the selection of desired data during data reduction workflows. This is achieved by automating the detection of when the laser has ablated through a sample using a smoothed time series, followed by removal of off-target data points. The functions are flexible and adjust dynamically to maximise the duration of the desired geochemical target signal, making this package applicable to a wide range of heterogenous bioarchives. Visualisation tools for manual validation are also included. CONCLUSIONS: LABLASTER increases transparency and repeatability by algorithmically identifying when the laser has either ablated fully through a sample or across a mineral boundary and is thus no longer documenting a geochemical signal associated with the desired sample. LABLASTER's focus on better data targeting means more accurate extraction of biological and geochemical signals.

How many specimens make a sufficient training set for automated three-dimensional feature extraction?

Deep learning has emerged as a robust tool for automating feature extraction from three-dimensional images, offering an efficient alternative to labour-intensive and potentially biased manual image segmentation methods. However, there has been limited exploration into the optimal training set sizes, including assessing whether artficial expansion by data augmentation can achieve consistent results in less time and how consistent these benefits are across different types of traits. In this study, we manually segmented 50 planktonic foraminifera specimens from the genus Menardella to determine the minimum number of training images required to produce accurate volumetric and shape data from internal and external structures. The results reveal unsurprisingly that deep learning models improve with a larger number of training images with eight specimens being required to achieve 95% accuracy. Furthermore, data augmentation can enhance network accuracy by up to 8.0%. Notably, predicting both volumetric and shape measurements for the internal structure poses a greater challenge compared with the external structure, owing to low contrast differences between different materials and increased geometric complexity. These results provide novel insight into optimal training set sizes for precise image segmentation of diverse traits and highlight the potential of data augmentation for enhancing multivariate feature extraction from three-dimensional images.