Sharing and Caring for the Bioarchaeological Heritage: What Should We Do With 3D Data in the Post-acquisition Stage?

Abstract

This paper discusses ways of collecting, storing and sharing 3D datasets of large archaeological assemblages, taking as a case study the Lepenski Vir anthropological collection. Several hundred human bones from individuals dated to Mesolithic-Neolithic have been digitalised through the use of image-based modelling, as well as volumetric and dedicated 3D scanners. The project was centred on providing undisturbed and meaningful options for accessing 3D scans of normally restrictively available physical samples, as well as establishing a base ground for further use of digital output. The results have been presented through the 3DHOP environment, providing a proxy to geometry and texture information, as well as contextual information about the finds. The authors elaborate on the workflow and storing strategies, as well as the possibilities for the public to interact with the digital catalogue. The paper also takes note of sensitivity of the digital 3D content created on the basis of archaeological record – and specifically human remains – for the process of online sharing.

1 Introduction

Lepenski Vir represents one of the most important sites for studying the process of Neolithisation in Europe and it is the eponym site of the local Mesolithic–Neolithic culture. Its position (in the north-central Balkans, on the border between Romania and Serbia, the Danube Gorges form a natural passage between south-eastern and central Europe), temporal continuity (Mesolithic–Neolithic occupation, ca. 9500–5500 calBC), the richness of its material culture (trapezoidal dwellings, raw material, ornaments, sculpted boulders, etc.) and human skeletal remains (>200) provides a unique contextual framework for examining the nature and dynamics of the Neolithic expansion, as well as the adaptation of the farming way of life to new natural and social environments.

Over the past decades, our knowledge about the life and the origins of the inhabitants of Lepenski Vir has greatly expanded. Intensive human occupation of this site started during the tenth millennium BC, suggesting that local Mesolithic hunter-fisher-gatherers (9500–6200 calBC) adopted a sedentary way of life prior to and independently of animal and plant husbandry, before the sixth millennium BC (Borić 2011; Dimitrijević et al. 2016). The stable isotope analysis and zooarchaeological record suggests that the Mesolithic diet was based primarily on the consumption of fish (freshwater and anadromous species) and hunted games(Dimitrijević et al. 2016; Živaljević 2017; Jovanović et al. 2019). During this period, we witness the appearance of dogs which were domesticated locally (Dimitrijević and Vuković 2015).

Radiocarbon dating models indicate a rapid expansion of the Neolithic in the Central Balkans at the end of the seventh millennium BC and confirm the pattern of population growth in the centuries following the spread of the first farming communities (Blagojević et al. 2017; de Becdelièvre 2020; Porčić et al. 2021). During this time (c. 6200–5950 calBC, Transformation/Early Neolithic period), intensive contacts between the inhabitants of Lepenski Vir and Early Neolithic communities from the neighbouring regions have been documented. This is apparent from the presence of new material elements characteristic of the Early Neolithic Stračevo culture at Lepenski Vir, including ceramics, polished stone axes, Balkan flint and shell beads made from Spondylus (Borić 2011). Moreover, unique artistic creations and architectural developments (anthropo-zoomorphic sculpted boulders, sophisticated trapezoidal buildings) also appear in the archaeological record. Elements of both continuity and novelty have also been observed in the mortuary repertoire during this period, such as Mesolithic-like supine positions and the practice of burying children under red-plastered floors of the buildings, pointing to the Southern Balkan and Anatolian Neolithic sphere (Borić and Stefanović 2004). Data from ancient DNA (Hofmanová 2016; Mathieson et al. 2018) and strontium radiogenic signals (Borić and Price 2013) indicate that at this time migrants came to Lepenski Vir who were genetically closer to individuals found at Early Neolithic sites in Anatolia. Palaeogenomic studies have also shown that local individuals with European-Mesolithic-like ancestry mixed with the newcomers from Anatolia (Hofmanová 2016; Mathieson et al. 2018; de Becdelièvre et al. 2020).

After 6000 BC, further important socio-cultural changes occurred at Lepenski Vir and this phase is marked as Early/Middle Neolithic period, c. 5950–5500 calBC. This period is characterised by the abandonment of the trapezoidal houses, the appearance of new type of domestic structures, domesticated animals (i.e., cattle, pig, goat and sheep) and inhumations in a flexed position, which are typical of Anatolian and European Early Neolithic communities (Borić 2013). Although domesticated animals appeared, stable isotope and zooarchaeological studies show that their consumption remained subsidiary to fish and wild game (Borić and Dimitrijević 2007; Dimitrijević 2008; Živaljević 2017; Jovanović et al. 2019). While most of the descendants of the local foragers continued to perpetuate their dietary traditions, the descendants of the migrants mostly adopted local fishing practices (de Becdelièvre et al. 2020). With regard to crops, a recent study suggests that some individuals dated to the Transformation/Early Neolithic and Early/Middle Neolithic phases may have consumed domesticated cereals (Filipović et al. 2017; Jovanović et al. 2021).

Although various elements of Neolithic culture arrived at Lepenski Vir, many studies showed that the transition from Mesolithic fisher-hunter-gatherers to Neolithic early pottery users was neither straightforward nor sudden, since many Mesolithic cultural traditions were still very much alive during the Neolithic. From a bioarchaeological perspective, this can be seen primarily in the continuity of the local economy, as well as mortuary practices that paint a picture of complex cultural syncretism at the onset of the Neolithic. Thus, Lepenski Vir gives valuable insights into the interactions between foragers and farmers along the Danube, in addition to highlighting the role of the social and ecological landscape that shaped the development of farming niches which may have subsequently further influenced the process of Neolithisation in the Balkans and Europe.

The 3D-scanning project was carried out by the combined effort of the Laboratory for Bioarchaeology, Faculty of Philosophy, University of Belgrade (http://bioarchlab.org/rs. Accessed 17 Jan 2021) and the Center for Biosystems, BioSense Institute, University of Novi Sad (https://biosens.rs/?page_id=12564&lang=en. Accessed 17 Jan 2021). So far, over 250 individual bones from grave contexts at Lepenski Vir and sites within the Danube Gorge have been scanned using multiple techniques. Image-based modelling was the main method of digitalisation used, followed by micro-computerised tomography (μCT) and computerised tomography (CT). The structured light scanner was least utilised, with access to the instrument being outsourced. For image-based modelling, Reality Capture was the chosen solution, the micro-computerised tomography was provided by BruknerSkyscan 1275 model and the dedicated structural light scanner unit was Range Vision: Spectrum. A medical CT scanner data,, was outsourced from a private medical laboratory (Fig. 9.1).

Fig. 9.1

The graph of the pipeline used for creating the 3D collection of Lepenski Vir

Two aspects should be singled out. First, neither of the research units involved allocated resources exclusively for 3D digitalisation, nor was this their primary work activity. Second, the acquisition of data was by and large completed within the relatively short time frame of two months of non-continuous scanning, with additional forays conducted on later occasions, mainly focused on μCT acquisition. The selection of samples included representatives of different age and sex groups, but the 3D scanning was not limited to diagnostic or impact-laden parts of the body. Rather, it covered every individual part that would be accounted for in a bioarchaeological record, regardless of size and fragmentation. Bias towards better preserved and more easily interpreted parts (by the public or by experts) would not have aligned with the idea of systematic coverage of the assemblage.

2 Sampling Strategy

The choice of remains to be entrusted to the 3D-scanning procedure was a multi-tier decision. At the core of the selection model are the questions driving the need to create digital replicas. We define them here as a group of major enquiries, followed by a set of secondary research questions to be resolved by the time the project is completed:

Is it possible to effectively share 3D content across a multitude of platforms with users who have varying levels of experience in dealing with 3D environments? The assumption is that a collection of 3D models can be distributed with context and paradata, in addition to geometric and (where applicable) colour information, while at the same time largely (but not completely) negating issues such as refresh rate (Molloy and Milić 2018), format interoperability and Level-of-Detail (LOD) thresholds.

In the knowledge that digital copies of human remains can be made accurate and with high levels of conformity to their physical counterparts, the project was interested in the prospect of treating the 3D-scanning deliverable as a tool for further studies. What is the procedure for making instruments out of 3D scans that can be used for comparative morphometric studies across the assemblage(s) and additive manufacturing?

The choice was thus guided by the need to represent all major sex and age categories, skeletal division groups and skeletal parts, with a particular focus on cranial skeletons (cranial skeleton examples, both complete and partially preserved, constitute nearly one-fifth of the digital replicas, however this is due to pronounced fragmentation of infant cranium bones). In addition, the presence of bone modification (i.e. disease-induced modification, traumatic events and slow processes) was identified as a trait to be fostered in the sample.

Secondary enquiries to be addressed by 3D scanning on the assemblage level include:

• How does a 3D scan fare as substitute to physical access to an anthropological sample?

  The assumption is that existence of a 3D collection of human remains from an (eponymous) archaeological site, would inspire engagement with the digital replicas on the part of students, researchers and the public, potentially reducing handling stress on actual (human) remains.

• Can the collection of 3D datasets covering a massive sample be conducted in the same dynamic way as any other classical form of archaeological documentation activity? The assumption is that, despite apparent logistical issues and unknowns (e.g. storing strategies, estimated speed of data production, lack of established standards), the 3D scanning of an assemblage of hundreds of ‘objects’ is a feasible secondary activity for a research unit with no primary interest in exclusive production of 3D content (just like any other type of documentation, such as image files or drawings of finds, that are created by mixed research teams).

• The greater the number of individual scans, the more prevalent issues relating to the storage and curation of progenitor data become. What should be kept for further use and what should discarded?

3 Providing 3D Models at a Large Scale

From the perspective of a craftsperson (as defined by Dolfini and Collins 2018), the process of 3D scanning can be made extremely elaborate. The number and position of scan stations, scanning or image loops around the artefact, exposure, overlap, speed or distance from the surface of interest can be varied experimentally, potentially leading to substantial improvements. However, there is not always time for such experiments for each object, as the issue is not just one of quality, but of quantity too. If there is more than one object, there will be more than one shape or material or both, and there will also be a timetable to be followed.

The author’s assumption was that dynamic development of software and hardware will make any detailed how-to guidelines covering each step obsolete or irrelevant, potentially after only a season or two, and there is thus little value in inscribing them in stone. More importance was attached to the means by which available instruments work with samples and to establishing general rules. For example, we found that the anthropological material handled contains fine details, which are faithfully reproduced in images and provide a multitude of features that can be recognised in the camera-network reconstruction stage of image-based modelling (Table 9.1). We also found that software can be guided to use fewer, but more accurately selected tie points (resulting in an overall better-quality model). In the case of volumetric scanning (Table 9.2), it was found that human teeth are particularly prone to the formation of ring artefacts (Boas and Fleischmann 2012) in the datasets and that this can effectively be eliminated by enabling random movement of the stage during acquisition. Moreover, even thin bones can be well documented by means of a structured light scanner (Table 9.3), with an increasing number of rotation steps during scanning, providing enough overlap between partially observed surfaces to successfully carry out the mutual co-registration stage. These (amicably designated) ‘hacks’ are signs of an emerging familiarity of the craftsperson with the material, and are quite similar to learning processes observed in standard methods of archaeological documentation.

Table 9.1 Data source – image-based modelling system of people of Lepenski Vir

Table 9.2 Data source – Micro CT system for people of Lepenski Vir

Table 9.3 Data origins – Structured light scanner for People of LepenskiVir

The technical pipeline manual was determined by means of a series of test trials and feedback between technician staff (with little or no training in anthropology) and anthropologists (with limited or non-existent experience of 3D scanning) as end users. This was done repeatedly for each source of datasets (i.e. image-based, volumetric and dedicated scanning systems), creating a general set of outlines that can be referred to each time a problem arises within the process. This would ultimately provide a highly detailed and accurate 3D reconstruction of the human remains from Lepenski Vir for each geometry source.

Once completed, 3D models were deployed through an open-source environment designed with the representation of cultural heritage in mind. The 3DHOP (Potenziani et al. 2015), produced by the Visual Computing Lab in Pisa, boasts well-stocked technical documentation that keeps pace with the setup phases (https://3dhop.net/. Accessed 17 Jan 2021). Included within the same environment are tools that are helpful for engaging with the assemblage, i.e. tools for the acquisition of metrical data, the orthographic projection of principal axial views, the enhancement of geometry characteristics via projected light and the possibility to create sections and isolate particular parts of the model (Fig. 9.2). The workflow forces a levelling of the playing field, converting diversified sources of 3D geometry into a single data-structure option. The stipulated format, multiresolution .nxs (.nxz), has no interoperability with industry-standard software, as it is pigeonholed into a single working domain. This makes it a prospective solution that allows for homogenous data structure, but at a hefty price of apparently losing advanced analytical potential available through other software. The strengths of 3DHOP and its standard are the ability to load and effectively display extremely high numbers of polygons, meaning that only the limits in consumption of storage space on the hosting structure determine the detail in which the 3D model can be shown.

Fig. 9.2

(a) standard visualisation of the model, (b) ortho-projection, (c) cross-section, (d) isolation of the part of the model by cross-section planes, (e) responsive lighting conditions for enhancing surface detail, (f) measurement tool at work

Each model was connected with a metadata spreadsheet. Attributes included are: information about the site and context, responsibility (i.e. the author of the data collection and the quality-verification personnel), the method used, the metrics of the acquired data (i.e. the image number for the image-based modelling and the scan number for dedicated scanning systems), the osteoarchaeological information (i.e. age, sex, skeletal division, skeletal part, particular bone), dating and publication reference (Fig. 9.3).

Fig. 9.3

(a) workbench of the catalogue with 3D models listed to the right and a search engine to the left, enabling visitor to browse by keyword, site, skeletal division, skeletal part, age category and sex of the individual, (b) workbench screen of an individual 3D object

3DHOP largely removes the visual appeal of a 3D scan out from the equation. The options for dramatisation of the digital content are limited, strictly focused on the dynamic lighting of the scenery; here we use dramatisation as a synonym for Physically Based Rendering (PBR), the photorealistic rendering of 3D content, or introducing emission, controlled transparency handles and the like. Whereas services such as Sketchfab excel at introducing multiple texture maps to increase the realistic look of the model or improving the model’s interaction with the digital environment and the public, 3DHOP relies on a spartan, function-driven design. This is not a shortcoming, but it does emphasise the innate expectations that archaeological finds could (should?) be striking to look at.

4 The 3D Collection of Lepenski Vir

People of Lepenski Vir (https://3d.biosense.rs/3DPortal/#/app-h/dashboard. Accessed 17 Jan 2021) is an online repository that provides limited access to the original files in the collection: there is no option to download either of the models made and a portion of the metadata is concealed from the visitor. The latter restriction was implemented to prevent the user panel from becoming overloaded with information that is not bioarchaeological (paradata related to the digitalisation process is not disclosed, but is viewable by an administrator). Admittedly, restricting direct access to the model stands in stark contrast to the democratisation brought about by the 3D-scanning process: the message delivered is that everyone can do a 3D scan, but there is no problem if only a few can review the product from bottom to top. The authors had relied on the online environment of 3DHOP to deliver options for assessing quality and context, but deliberately omitted the potential for reuse, since it was unclear what the afterlife of the generated models would be if these were fully downloadable. One could evoke the list of industries that use 3D scanning (e.g. interior and exterior design, gaming, product placement, etc.) and suggest that most users would be liable to find a secondary purpose for a 3D model coming from an archaeological assemblage. Moreover, ethical issues would not be far behind. While I do not believe that making Lepenski Vir models freely downloadable would invite the massive and uncontrolled repurposing of the assemblage, there should be a comprehensive environment that protects all involved: creators, content and stakeholders. It was proposed that the feasibility of open access to 3D-scanned human remains is governed by two factors: the temporal gap and the culturally sensitive nature of the material. Recent human remains are more liable to be withheld from public access (and even more so those originating from modern conflicts), while those of ancient origin, especially of high importance to the research community, are easier to distribute. The cultural sensitivity of the local living communities, as important stakeholders who claim the ownership over displayed and managed objects, represents the second factor (Ulguim 2018). Concerns about the misuse of the digital osteoarchaeological record from Lepenski Vir arise from both perspectives, as a rather intimate connection with the country’s population is established through the great age and scientific importance of the assemblage. Lepenski Vir is strongly present in the publicly constructed archaeological landscape of the Republic of Serbia, as only a handful of other monuments of (archaeological) cultural property exist. It is also a target for a number of pseudo-archaeological agendas, often driven by misplaced sensationalism (Milosavljević and Palavestra 2017) and a desire to find ever older roots for the nation in the ancient past.

Furthermore, the local legislative framework recognises human remains of ancient populations as the cultural property of the state, and the 3D models of the remains are loosely viewed as documentation pertaining to the original find, and thus, broadly interpreted, as taking on characteristics of the cultural property itself. Hence, facilitating full access to the models, while imposing a strict view on ownership and licensing, without the tools to claim and protect these asserted rights in the digital environment would likely not give the desired result.

That said, none of the above considerations outweighs the benefits of direct reuse of 3D models to foster engagement between the community and bioarchaeological assemblages, or a re-evaluation of the material through research activities. What 3D documented archaeological object excels at is the sheer number of new reinterpretations which it enables: a 3D model can be enhanced, with regard to both geometry and colour information, materialised beyond the basic metrics into a tool for direct comparison, or introduced into a digital environment for immersive handling.

From this perspective, the tools provided with the Lepenski Vir catalogue are limited, but point ahead to the next level of development. An overhaul of the catalogue would ideally include the possibility to download a progenitor 3D object (not the nexus standard) and increased contextualisation through the introduction of engagement handles on the model that provide detailed information about points of interest. Such features would serve to prevent the 3D model from being relegated to the role of a mere showcase and allow it to be more actively used. Notably, the introduction of persistent identifiers for each 3D object would allow for direct referencing (Ulguim 2018), provided that the digital repository is a stable source of data flow for a longer period. A prerequisite for this would be the formulation of a clearer legal framework for dealing with 3D models of cultural heritage.

The project explored the option of creating instruments from 3D scans to facilitate further uses, beyond the inclusive access to digital content, for purposes related to morphometric comparative studies and additive manufacture. Data complexity was identified as one of the primary bottlenecks impeding effective manipulation of scanned human remains. Detail-heavy models are also massive datasets, and in virtue of the composition of collection, models are units belonging to a multi-part system: an arm, leg, pelvis, shoulder, chest or cranial group of bones, which, in turn, can be made to work together in the digital realm. Whereas the 3DHOP partially handles the issue of casual/fast browsing of the 3D content in a way similar to browsing images on a hard drive, there is an interest in introducing direct model-to-model shape and volume comparison. This interest has been also expressed by researchers outside the project staff, following the first presentation of the Lepenski Vir 3D collection in EAA Bern 2019. No viable solution for this in 3DHOP architecture has so far been devised for the Lepenski Vir project: a separate piece of software, such as Cloud Compare or Meshlab, can be used externally, but the procedure can be burdened by surface mesh density, available hardware and user familiarity with the software. It also requires direct access to progenitor 3D models. In order to facilitate the comparative potential in an assemblage, uniform scale needs to be imposed on the data. The volumetric and dedicated scanners used achieve this through calibration at the pre-scan stage, but image-based modelling require that the scenery contains points with known distances, preferably with only minute errors in dimensions. Colour information can be made uniform across the sample by colour checkers, however this aspect can vary between acquisition platforms (image-based modelling/dedicated scanners), as well as between models, where light-source distance and resolving power can affect the exposure and acquired dynamic range of images (image-based modelling).

When it comes to additive manufacture, the geometrical integrity of the models is paramount. Printing human remains requires that the appearances of objects without thickness, disconnected vertices and edges, internal faces, etc. (commonly known as non-manifold geometry) be removed from the surface model. These may appear at multiple stages, as the result of decimation of the model, closing gaps in the object surface or straight from the software/hardware used to acquire data. Once again, optimisation of the models is an important factor: an overly detailed model will still be printed within the limits of the printer’s technical capacity, but excessive complexity (which will anyways be omitted from printed replicas) may cause holdups in laying out the printing procedure. The orientation of the digital model is also of importance for certain printing procedures – polylactic filaments (PLA) are layered prints, and the orientation of the layers (printing position of the object) may affect rendering of surface features.

Assessments of 3D scans – regardless of whether they are presented through a 3DHOP environment or third-party software – as a suitable replacement for direct access to the physical object revealed overlaps between experiences. The reported loss of synchronisation in working with a digital copy was noticed, referring to 3D model manipulation and coordinate systems of the 3D model (what is right and left, up and down, and how to facilitate desired orientation), as well as limited depersonalisation of handling and separation between the physical and digital. The latter term refers to the reported notion that the digital version is a completely new object and that repeated visits to the physical sample were needed to establish a connection with it. However, the surface and (where available) internal features of the remains – notably trauma- and disease-induced changes (for comparison download access to partial or complete 3D models was made here, here and here) – translate well into a new environment and are identifiable and accountable, with clear morphometric traits. It is, however, still unclear whether the disparity in experience is removed after prolonged exposure to the dataset working domain.

The Lepenski Vir project has provided a convincing argument that 3D scanning can be performed in the background of other research activities at a physical anthropology laboratory (or indeed any other functional research unit). Whereas volumetric model creation is limited to inducted and certified technicians, other close-range 3D scanning techniques are taught and upgraded by stable and continuous submission to the process, even for users with no background in any form of digitalisation. Of special note is that the applicant should be encouraged to engage in experimental forays, as these actively build a sense for the potential of the heritage 3D content and the capacity to manage and analyse it (thus reducing the chances that this content will sit out its time on a backdrop hard drive). Unsurprisingly, however, the speed achieved does not compare to the speed with which more classical forms of archaeological documentation are created. The People of Lepenski Vir project has shown that, in one working shift, the raw data needed to make a complete 3D model can be created for up to 10 samples from the collection. When processing was taken into account, however, our numbers went down to 3–5 fully completed samples per day. Furthermore, given the fact that 3D models are not able to provide instant information, but rather require additional efforts, tackling a body of finds without access to a streamlined workflow can result in a protracted entanglement, where information and effort is piled up, but no quality feedback or deliverables are produced.

This brings us to the last issue, concerning the storage of raw and intermediate products. The People of Lepenski Vir project preserved the core products and outputs, discarding most of the transitional steps between the two. This is exemplified by all three data sources, in which dependencies, offshoot products and temporary datasets were made, which were not vital to the continued life of the project or individual scan. Arguably, there could be benefits to being able to respond quickly and retrace one’s steps to a particular moment in the workflow. However, the general sentiment on the Lepenski Vir project was that this would not be an optimal investment of time and resources. Instead, a record was kept of how models were generated, from the acquisition stage to the final output. This largely conformed with the point 7 of the ‘Basic Principles and Tips for 3D Digitisation of Cultural Heritage’ (https://ec.europa.eu/digital-single-market/en/news/basic-principles-and-tips-3d-digitisation-cultural-heritage#2.%20Select%20what%20to%20digitise. Accessed 20 Jan 2021). We expect that the processing software will continue to make major jumps and that old datasets can be revisited if the opportunity arises to obtain improved results.

The online catalogue represents a decent step forward in helping make the bioarchaeological assemblage of Lepenski Vir accessible to observers. It appears that the community of researchers concerned with 3D documentation and content in archaeology and cultural heritage has reached at a stage where both the curation and distribution of 3D is branching out through experiments with reliable and widely compatible solutions to achieve both goals (Erolin et al. 2017; Fanini et al. 2019). While it seems that this is leading to the formation of individual clusters of context-specific collections, important work is being done and the availability of the 3D output, for both the general and expert public, is improving steadily, ultimately moving towards formalisation.