"Reconstructing fragmented bones, reconstructing fragmented lives"
Human skeletal remains can offer key information on diverse aspects of past life but, like all archaeological materials, they only give a glimpse into the life of past individuals; hence, a fragmented perspective into our ancestors’ lives. Their interpretative potential is greatly inhibited by the fact that such remains are often found fragmented due to several anthropogenic and natural taphonomic agents, such as funerary treatment, animal activity, soil pressure and others. This fragmentation severely limits the information that may be extracted from human bones in terms of metric analysis, geometry and morphology. RECONSTRUCT aims at producing 3D morphable models for the main elements of the lower and upper limbs of the human skeleton, which will be then used to infer the missing morphology of fragmented or incomplete bones.
In order to achieve this aim and address a major current limitation in human skeletal analysis, RECONSTRUCT will integrate approaches from osteoarchaeology, forensic anthropology, biology, engineering, and data science. The results of the project will maximize the information that may be extracted from bioarchaeological research, contributing to a more comprehensive assessment of past life parameters. RECONSTRUCT will also have major implications in forensic anthropology since the study of modern skeletal remains suffers from similar limitations in terms of partial preservation as those witnessed in archaeological bones. Finally, the source code and raw data emanating from RECONSTRUCT will be made open access, greatly enhancing the extension of this approach to zooarchaeology, palaeoanthropology, and the prosthetic implants industry.
Data collection
RECONSTRUCT’s success relies on a very large assemblage of human long bones, which have been sourced from various skeletal reference collections across Europe and have been digitized by means of 3D photogrammetry to produce the largest virtual skeletal collection to date. Skeletal material was used from the Athens Collection, which is housed at the Department of Animal and Human Physiology of the National and Kapodistrian University of Athens, the Cretan Collection, which is housed at the Forensic Medicine Unit of the Faculty of Medicine of the University of Crete, and the Granada Collection, which comprises individuals from the Cemeterio de San José Cemeterio in Granada, Andalucia and it is currently curated at the Department of Toxicology and Physical Anthropology of the University of Granada. In total, 3618 skeletal elements of the lower and upper limbs of 518 adult individuals have been digitized for the purposes of the RECONSTRUCT project. The digital twins have been openly shared with the curators of the respective collections, so that they are actively used by the universities housing these collections, but also by external scholars, for further research. Given the high accuracy of these models, they can effectively replace the direct handling of the skeletal collections for macroscopic analyses, thus, they can contribute to more ethical anthropological research practices.
A direct derivative of the aforementioned collection of digital twins is the public dataset of cross-sectional geometric properties of human long bones, which is licensed under Creative Commons Attribution 4.0 International. This dataset is also available as a GNU Octave package named csg-dataset. This dataset contains 61 measurements extracted from 3618 digital twins of intact femora, humeri, tibiae, and ulnae, which belong to 518 adult individuals of known sex and age-at-death, derived from the aforementioned European skeletal reference collections. The dataset also contains the cross-sectional contours from which the aforementioned measurements have been calculated.
3D Morphable Bone Models
A significant milestone towards accurate surface reconstruction of fragmented bones is the development of 3D Morphable Bone Models (3DMBMs), which can be through as statistical representation of the complete surface of the skeletal element comprising the variation observed within a population sample. The RECONSTRUCT project has utilized the large dataset of digital twins of the aforementioned skeletal collections in order to train 3DMBMs for the femur, humerus, tibia, and ulna bones. In practical terms, 3DMBMs utilize prior knowledge of the shape of an assemblage of skeletal elements and use it as a basis to represent or generate new plausible skeletal elements of the same type. From a statistical perspective, a 3DMBM comprises a large covariance matrix, which is used to modulate random generation from a multivariate distribution accurately describing the group of objects it was trained on.
Femoral proximal epiphysis
The animation on the left comprises rendered images of several 3D models of the femoral proximal epiphysis, which have been auto-generated from the 3DMBM trained on available femurs. The visualized 3D models do not necessarily coincide with any of the skeletal elements available in the training sample, but their surface accurately depicts plausible shapes of the femoral proximal epiphysis from south European populations. This range of plausible shapes is exactly what the 3DMBMs aim to capture and statistically describe. In more practical terms, this means that (almost) any non-pathological variation observed in the femur of south European populations can be represented and/or regenerated by the underlying multivariate distribution of the trained 3D Morphable Bone Model.
Similarly, the animation on the right illustrates the corresponding distal epiphyses of the same auto-generated artificial femurs, whose proximal epiphyses are displayed above. The correspondence of the two epiphyses as illustrated in these two animations demonstrates the capacity and usefulness of 3DMBMs in missing surface reconstruction. The rationale behind this approach is that the variation of morphological features in bones that are non affected by pathological conditions uniformly affects the entire surface of the bone. Consequently, being able to match the surface of a partial bone to a particular instance of the 3DMBM allows for the complete reconstruction of the missing surface in a much more accurate way that a simple bone atlas could ever achieve.
Femoral distal epiphysis
More updates comming soon!
The "RECONSTRUCT" project has been funded by the European Union's Horizon Europe research and innovation program under the Marie Skłodowska-Curie Actions granted to Andreas Bertsatos [Program/Call: HORIZON-MSCA-2022-PF-01, Proposal: 101104702-RECONSTRUCT]. It was conducted at the STARC of The Cyprus Institute under the supervision of Ass. Prof. Efthymia Nikita.
