Quantifying the Benefits of Imputation over QSAR Methods in Toxicology Data Modeling.

Imputation machine learning (ML) surpasses traditional approaches in modeling toxicity data. The method was tested on an open-source data set comprising approximately 2500 ingredients with limited in vitro and in vivo data obtained from the OECD QSAR Toolbox. By leveraging the relationships between different toxicological end points, imputation extracts more valuable information from each data point compared to well-established single end point methods, such as ML-based Quantitative Structure Activity Relationship (QSAR) approaches, providing a final improvement of up to around 0.2 in the coefficient of determination. A significant aspect of this methodology is its resilience to the inclusion of extraneous chemical or experimental data. While additional data typically introduces a considerable level of noise and can hinder performance of single end point QSAR modeling, imputation models remain unaffected. This implies a reduction in the need for laborious manual preprocessing tasks such as feature selection, thereby making data preparation for ML analysis more efficient. This successful test, conducted on open-source data, validates the efficacy of imputation approaches in toxicity data analysis. This work opens the way for applying similar methods to other types of sparse toxicological data matrices, and so we discuss the development of regulatory authority guidelines to accept imputation models, a key aspect for the wider adoption of these methods.

GAKTpore: Stereological Characterisation Methods for Porous Foams in Biomedical Applications.

In tissue engineering, scaffolds are a key component that possess a highly elaborate pore structure. Careful characterisation of such porous structures enables the prediction of a variety of large-scale biological responses. In this work, a rapid, efficient, and accurate methodology for 2D bulk porous structure analysis is proposed. The algorithm, "GAKTpore", creates a morphology map allowing quantification and visualisation of spatial feature variation. The software achieves 99.6% and 99.1% mean accuracy for pore diameter and shape factor identification, respectively. There are two main algorithm novelties within this work: (1) feature-dependant homogeneity map; (2) a new waviness function providing insights into the convexity/concavity of pores, important for understanding the influence on cell adhesion and proliferation. The algorithm is applied to foam structures, providing a full characterisation of a 10 mm diameter SEM micrograph (14,784 × 14,915 px) with 190,249 pores in ~9 min and has elucidated new insights into collagen scaffold formation by relating microstructural formation to the bulk formation environment. This novel porosity characterisation algorithm demonstrates its versatility, where accuracy, repeatability, and time are paramount. Thus, GAKTpore offers enormous potential to optimise and enhance scaffolds within tissue engineering.

2D single crystal Bragg-dip mapping by time-of-flight energy-resolved neutron imaging on IMAT@ISIS.

The cold neutron imaging and diffraction instrument IMAT, at the second target station of the pulsed neutron and muon source ISIS, is used to investigate bulk mosaicity within as-cast single crystal CMSX-4 and CMSX-10 Ni-base superalloys. Within this study, neutron transmission spectrum is recorded by each pixel within the microchannel plate image detector. The movement of the lowest transmission wavelength within a specified Bragg-dip for each pixel is tracked. The resultant Bragg-dip shifting has enabled crystallographic orientation mapping of bulk single crystal specimens with good spatial resolution. The total acquisition time required to collect sufficient statistics for each test is ~ 3 h. In this work, the influence of a change in bulk solidification conditions on the variation in single crystal mosaicity was investigated. Misorientation of the (001) crystallographic plane has been visualised and a new spiral twisting solidification phenomena observed. This proof of concept work establishes time-of-flight energy-resolved neutron imaging as a fundamental characterisation tool for understanding and visualising mosaicity within metallic single crystals and provides the foundation for post-mortem deduction of the shape of the solid/liquid isotherm.

Automatic Recognition of Dendritic Solidification Structures: DenMap.

Dendrites are the predominant solidification structures in directionally solidified alloys and control the maximum length scale for segregation. The conventional industrial method for identification of dendrite cores and primary dendrite spacing is performed by time-consuming laborious manual measurement. In this work we developed a novel DenMap image processing and pattern recognition algorithm to identify dendritic cores. Systematic row scan with a specially selected template image over an image of interest is applied via a normalised cross-correlation algorithm. The DenMap algorithm locates the exact dendritic core position with a 98% accuracy for a batch of SEM images of typical as-cast CMSX-4® microstructures in under 90 s per image. Such accuracy is achieved due to a sequence of specially selected image pre-processing methods. Coupled with statistical analysis the model has the potential to gather large quantities of structural data accurately and rapidly, allowing for optimisation and quality control of industrial processes to improve mechanical and creep performance of materials.