Cremation burials present particular analytical challenges due to the intrinsic fragmentation of burned bone and further fragmentation during excavation and processing. Over recent decades, radiographic imaging—particularly CT—has been used to non-destructively inspect urned cremations to locate remains, identify inclusions, and guide excavation. Technological improvements in scanners and software have increased the information retrievable, but substantial obstacles remain when working with urned cremations: overlapping density values of different materials, scan artifacts, limited spatial resolution for larger vessels, and the high cost or impracticality of higher-resolution alternatives such as micro-CT or synchrotron imaging.
This exploratory study tested a relatively simple and time-efficient virtual workflow applied to average-quality tomographic images with the explicit aims of extracting and assessing qualitative and quantitative information about:
The virtual results were systematically compared with subsequent physical micro-excavation and osteological analysis to evaluate accuracy, limitations, and the added value of the two approaches combined.
The sample comprised 10 non-micro-excavated urns recovered from the Middle–Late Bronze Age necropolis of Vicofertile (Parma, Italy). Each urn was scanned using conventional Computerised Tomography (CT). Following imaging, all urns underwent micro-excavation and full physical osteological analysis. The study therefore provides paired virtual and physical datasets for direct comparison.
The authors describe the rationale for using conventional CT: jar size often precludes micro-CT scanning, and alternatives such as synchrotron or specialised facilities may be costly or inaccessible. Given these constraints, the study evaluates what information can be retrieved from readily available imaging technology and how a relatively simple segmentation strategy might support exploratory analyses.
A central challenge in virtual analysis of burned remains is segmentation—the process of assigning voxels in the tomographic volume to distinct materials based on density (Hounsfield units). In urned cremations, bone, soil, ceramics, and porous materials often have overlapping grey values, complicating automated discrimination and making segmentation time-consuming. The authors applied a practical segmentation strategy tailored to average-quality CT images intended to be less complex than previously published semi-automated frameworks.
Where artifacts or density overlaps occurred, manual corrections were performed. The workflow emphasised extraction of qualitative markers (e.g., visible fractures, bone clustering) and coarse quantitative metrics (e.g., counts of distinguishable anatomical regions, volumetric estimates) that could be compared against results from the physical analysis.
Key outcomes from comparing CT-based observations with physical analyses included:
Age classes and burial taphonomy: Data inferred from the scans about age categories and taphonomic indicators were confirmed by physical osteological analysis.
Thermal alterations: Heat-induced changes visible on scans—such as fractures associated with burning—were corroborated by the excavation and physical inspection of bones.
Minimum Number of Individuals (MNI): The MNI hypothesised from CT scans was 10 across the sample; after micro-excavation, the MNI increased by one, indicating at least one instance where CT underestimated individual counts.
Sex estimation: Sex could not be reliably estimated from the CT images at the resolution and image quality available in this study.
These results indicate that some categories of information (age class, taphonomy, thermal damage) can be fairly robustly predicted from conventional CT in many cases, while other metrics (sex, precise MNI) remain difficult without physical analysis or higher-resolution imaging.
The study evaluated the vertical and horizontal distribution of anatomical categories within each urn, dividing elements into cranial, long bone, spongy bone, upper limb, lower limb, and trunk categories. Among these, the vertical cranial distribution was the most accurately predicted from CT images when compared with micro-excavation results. Other anatomical categories showed lower accuracy in virtual prediction, likely due to fragmentation, overlapping densities, and partial volume effects in average-quality CT scans.
The authors used the segmentation outputs to map apparent clusters and layering of skeletal material within vessels as an aid to understanding depositional processes and taphonomy, but emphasize that these spatial interpretations required physical validation.
Several limitations of the CT-based approach are highlighted by the study and by prior work discussed in the introduction:
Density overlap and grey-value similarity between materials within urns hinder automatic or semi-automatic segmentation and increase the need for manual correction.
Artifacts and limited resolution in conventional CT scans reduce the ability to resolve small or highly fragmented elements, affecting estimations of MNI and anatomical identification.
Jar size often precludes use of micro-CT, which would provide higher spatial accuracy; alternatives such as synchrotron facilities or specialised industrial scanners are costly and not always available.
The proposed segmentation strategy is intentionally simpler and more accessible than some complex semi-automated frameworks, but it does not eliminate the requirement for physical excavation and manual verification in many cases.
Overall, the physical analyses proved essential to refine virtual results, compensate for low-quality images, and recover data that were ambiguous or missing from CT alone.
This work represents a systematic comparison between tomographic imaging and physical analysis of urned cremations at a scale larger than many previous studies. It demonstrates that conventional CT can yield meaningful information about burial taphonomy, age classes, and heat-induced bone alterations, and can reasonably predict certain spatial patterns such as vertical cranial placement. However, CT underestimated MNI in at least one case and could not support sex estimation at the image quality used.
The authors conclude that CT-based exploratory analysis is a valuable non-destructive first step that can guide laboratory strategies and excavations, but physical micro-excavation remains necessary to validate and expand virtual findings. Future improvements could include higher-resolution scanning where possible, development of more robust segmentation algorithms tailored to cremation assemblages, and protocols that integrate virtual and physical workflows to maximise information recovery while minimising unnecessary handling.
The data and scripts supporting this study are openly available in Zenodo (DOI: 10.5281/zenodo.15042290). The research is part of the corresponding author’s PhD project and received funding from several scholarships and travel grants; the funders had no role in study design, data collection and analysis, decision to publish, or manuscript preparation. The authors declare no competing interests.