(a) Intra- and inter-observer measuring repeatability.

To confirm that the semi-automated measuring protocol of VERA 2.0 is repeatable within and among different observers (following a thorough and meticulous self-study of all details included in the step-by-step protocol), repeatability tests were conducted on three enthesis-bearing bone structures of major functional importance that correspond to three of the most frequently studied anatomical regions of the postcranial skeleton: the arm (humerus), the hand (thumb metacarpal), and the foot (calcaneus). These consisted of the insertion site of opponens pollicis (OP), the deltoid tuberosity, and the calcaneal tuberosity (i.e., the attachment structure of the Achilles tendon, which is often considered as the archetypical fibrocartilaginous enthesis). The intra-observer repetitions were performed by the author, while inter-observer repetitions involved a different observer for each of the three enthesis-bearing bone structures (for more information on the three additional observers, see Acknowledgements). A total of 10 bones were measured in each case, including specimens from both anatomical sides (Table 1). All repetitions were done independently and on different days, while training only included a single demonstration of the protocol, including a clear description of the SP and BSR for these enthesis-bearing structures (Figs 2 and 3; also see detailed information on these entheses in the protocol). Following statistical recommendations for testing precision in entheseal 3D areas [26], repeatability was evaluated using the Lin’s concordance correlation coefficient (Lin's CCC), which showed values of 0.95 or above, reflecting a strong level of repeatability and reproducibility [43,44]. Additionally, a series of paired t-tests found no significant differences (P >  0.05) between repetitions within the same observer (the author) and across observers. Despite the high repeatability observed in this study, as noted in previous methodological work on VERA 1.0 [17,27], morphological complexity varies significantly across different enthesis-bearing bone structures of the body, inevitably leading to different levels of difficulty in selecting an identical BSR across individuals. Therefore, I strongly recommend that future users of VERA 2.0 always statistically confirm their measuring repeatability before conducting analyses, as was also suggested for VERA 1.0 [3,26,27]. Users should follow the step-by-step protocol carefully, ensuring that both SP and BSR are as clearly defined and consistent across their sample as possible. If any practical difficulties arise, users are always encouraged to contact the developer for additional guidance or support.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Results of the repeatability tests.

https://doi.org/10.1371/journal.pone.0321479.t001

(b) Experimental validation using laboratory animals.

An original analysis was conducted to evaluate the reliability of the VERA 2.0 method for reconstructing habitual activity based on the morphological comparison of enthesis-bearing bone structures. The sample consisted of 15 laboratory-raised Eastern wild turkeys, all female and one-year-old, previously analyzed in a study employing the traditional VERA 1.0 approach [14]. During the experiment, turkeys were divided into three groups: uphill runners, downhill runners, and controls, with treadmill running over 10 weeks (at a speed of 2.5 m s^− 1, for 30 min a day, 4 days a week). The original VERA 1.0 analysis had identified clear patterns in three distinctive enthesis-bearing structures of the femur (associated with muscles gluteus primus, medial gastrocnemius, vastus medialis, and adductor magnus), clearly distinguishing runners from controls. In this study, I applied VERA 2.0’s semi-automated protocols to the same samples and entheses. Nevertheless, because the birds’ 3D surface models were relatively small and already quite smooth, applying the smoothing step and using a 2 mm threshold for artifact removal consistently resulted in “0” areas being identified across most of the sample’s specimens. Therefore, it was necessary to skip the initial smoothing step and reduce the artifact removal threshold to 1 mm, following the protocol’s recommendations (see S1 File). It is important to note that this issue did not arise in the human skeletal analyses discussed in the next subsection (and shown in Figs 1–3), which also included relatively small hand bones.

Subsequently, a principal component analysis (PCA) and other multivariate models were used to compare the groups (following [26,45]). The results (Fig 4) show clear differences between controls and both groups of runners, along with a trend distinguishing uphill from downhill runners. In the figure’s side images (left and right), examples of delineated irregularities are provided for each enthesis-bearing structure: the BSR is highlighted in red, while the blue areas indicate the entheseal changes identified. These examples illustrate the patterns of controls (negative PC1 scores) versus runners (positive PC1 scores). Importantly, including “strides” (the exact number of steps each bird took) in the PCA revealed a strong covariance between the identified 3D surface changes (in mm²) and the exact number of strides performed (Fig 4). Three additional analyses were conducted in R, to explore this pattern [45] (Table 2). A multivariate analysis of variance (MANOVA) comparing groups (without the strides variable) showed significant and very strong differences (Pillai’s Trace =  0.88). A multiple regression model showed that the identified surface changes significantly predicted the number of strides, explaining 52% of their total variance (based on the R² value). Finally, an analysis of covariance (ANCOVA) adjusting for body mass confirmed that the observed entheseal changes were not due to body size differences among birds. For all three analyses, the necessary statistical assumptions were met, including linearity (but no multicollinearity), absence of outliers (in the variables and the residuals), approximate normality of residuals, homogeneity of slopes, homoscedasticity per group, and homogeneity of variance-covariance matrices, in line with standard statistical recommendations [26,43,45,46]. Overall, these results confirm VERA 2.0’s reliability in reconstructing activity in experimental settings, without necessarily needing size adjustment. In total, measuring and analyzing the surfaces of all 45 enthesis-bearing bone structures required one working day.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of statistical results from three analyses comparing control, uphill, and downhill runner turkeys based on 3D surface changes on enthesis-bearing bone structures, measured using VERA 2.0.

https://doi.org/10.1371/journal.pone.0321479.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Principal component analysis (PCA) results comparing control turkeys with uphill and downhill runners, based on 3D surface changes on enthesis-bearing structures, measured using VERA 2.0.

In the side example images (representing the entheseal patterns of negative and positive PC1 scores, respectively), BSR regions are highlighted in red, with blue areas showing the automatically defined surface irregularities.

https://doi.org/10.1371/journal.pone.0321479.g004

(c) Validation using documented human skeletons.

To further validate the reliability of the VERA 2.0 protocols for reconstructing physical activity in humans, I conducted a study using anthropological remains from the 19th-century Basel-Spitalfriedhof collection, curated by the Natural History Museum of Basel, Switzerland [18,47,48]. This collection is uniquely valuable for biological anthropology due to its rare, detailed documentation on each individual’s occupation, duration of employment, occupational changes during life, and exact work hierarchy [18,47]. Previous studies using VERA 1.0 demonstrated clear distinctions between occupational activity levels in biological males, separating heavy manual laborers (mainly construction workers, including stonemasons, carpenters, and bricklayers) from precision workers engaged in low-intensity tasks (e.g., tailors, shoemakers, and painters) [18]. Similarly, research on biological females from the same collection identified distinct patterns in females with specific occupations (“FSO”), such as trained tailors and seamstresses, compared to those with more generalized tasks (e.g., housemaids and factory workers without specifically archived job descriptions) [19].

In this validation study, I re-measured selective key enthesis-bearing bone regions from these individuals using VERA 2.0’s semi-automated process. The resulting measurements and exact specimen labels are available open-access in the Zenodo repository (https://doi.org/10.5281/zenodo.15000406). All necessary permits were obtained for the described study from the responsible institution (Natural History Museum of Basel, Switzerland), which complied with all relevant regulations.

Without applying any size adjustments, Fig 5 shows that simply plotting the 3D surface size of irregularities within bone regions associated with two muscles – OP and extensor pollicis brevis – on a bivariate XY plot visually separates long-term construction workers from precision workers. This separation represents entheses of muscles associated with thumb extension and opposition. The figure includes side images of each occupational group’s enthesis-bearing bone regions, with SP and BSR highlighted in red and surface irregularities in blue. Similarly, the female plot (Fig 6) of two enthesis-bearing bone regions (related with muscles OP and opponens digiti minimi) demonstrated a clear separation between FSOs and those with generalized tasks. FSOs showed significantly larger 3D changes in both regions, which represent muscles used for thumb and little finger opposition. To statistically support these findings, I additionally performed a MANOVA to compare between occupational trends in each case (males and females). Significant differences between occupational groups (P <  0.05) were found in both sexes (Table 3), while all assumptions for MANOVA were met [26,43,45,46]. Spearman’s correlation tests were also conducted to evaluate the relationship between the observed surface irregularities and factors such as biological age and body size (using femoral head diameters [49]), finding no significant correlations (P > 0.05) in this well-documented sample [26].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multivariate analysis of variance (MANOVA) results comparing the two occupational groups within each biological sex.

https://doi.org/10.1371/journal.pone.0321479.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Bivariate plot showing variation between biological male construction workers (heavy manual laborers) and precision workers, based on 3D surface regularities within bone regions associated with muscles opponens pollicis (OP) and extensor pollicis brevis (EPB).

Larger surface irregularities were observed in construction workers, highlighting their increased biomechanical stress due to manual labor. Side images show examples of the BSR (highlighted in red) and surface irregularities (in blue) for each occupational group.

https://doi.org/10.1371/journal.pone.0321479.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Bivariate plot showing the distinction between female workers with specifically defined occupations (FSOs) and generalized workers (mainly maids and factory workers), based on 3D surface regularities within bone regions associated with muscles opponens pollicis (OP) and opponens digiti minimi (ODM).

FSOs exhibited overall larger surface irregularities, likely reflecting more biomechanical stress in occupational tasks requiring coordination of the thumb and the little finger via opposition (in line with previous results and activity interpretations regarding these individuals [19]). Side images show examples of the BSR (red) and surface irregularities (blue) for each group.

https://doi.org/10.1371/journal.pone.0321479.g006

Interestingly, both male heavy manual laborers and FSOs tended to display significantly larger surface irregularities in both attachment sites analyzed, a result different from VERA 1.0, which found distinct occupational differences in the proportions among different enthesis-bearing structures (rather than larger raw values in one of the groups compared). This is probably because VERA 1.0 quantifies the entire supporting bone structures associated with an enthesis (e.g., the 3D surface of the entire tubercle within which the tendon footprint is located), whereas VERA 2.0 aims to filter out discrete surface irregularities (“changes”) within that area (see discussion in [16]). Based on the literature, the extent of these localized bone alterations is presumably more dependent on the amount of biomechanical stress directly applied on the surface by the attaching soft tissue [1,4,5,10,13,14,16,50] compared to the bone remodeling responses that affect the gradual uniform change of the entire supporting bone structure (e.g., the tubercle) [9]. This scenario would explain why the more strenuous occupational trends within males or females presented overall greater changes in both enthesis-bearing structures analyzed (for the strenuous manual activities of FSO individuals, see discussion in [19]), which was also the case for the habitual runners of the experimental animal dataset (Fig 4). It would also explain why, in contrast to VERA 1.0’s raw measurements (prior to adjustment), VERA 2.0’s raw measurements do not significantly correlate with body size or age and can directly distinguish occupational trends without requiring size adjustment (at least in the datasets of the present study). This is because, in case these entheseal surface irregularities are primarily stress-inflicted, they would not be expected to be as correlated with body size as larger bone structures that first form during development [3,26,12]).

However, this does not mean that analyzing multivariate proportions among enthesis-bearing structures is irrelevant for VERA 2.0 data. Different activity patterns engage various muscles, and those used more frequently are expected to exert greater loading onto their associated bone regions (e.g., as seen in the varying clarity of occupational differences between OP and the more distinct opponens digiti minimi in Fig 6). This proportional approach may better distinguish occupational trends than analyzing individual enthesis-bearing bone structures in isolation [3,15,12]. Thus, the integrative univariate and multivariate statistical analyses outlined in VERA 1.0’s formal protocol [26] remain highly recommended for VERA 2.0 measurements.

(d) A pilot anatomical examination and future research possibilities.

The presence of various surface irregularities on enthesis-bearing bone structures, often simplified in the anthropological literature as “entheseal changes”, has been traditionally associated with cumulative biomechanical loading inflicted by the attaching muscles, tendons, or ligaments (e.g., [1,2,5,8,10]). Multiple anatomical and clinical studies have studied these traits from a histological perspective, explicitly addressing the factors driving the appearance of bony projections within entheseal footprints (i.e., bony spurs, lipping, and exostoses) [35–37,13,51–53]. However, no research has utilized fresh cadaveric dissections to verify the exact correspondence between soft tissue and the specific types of scored ‘entheseal changes’ commonly analyzed in anthropological research to infer physical activity in bioarchaeological contexts (i.e., ranked stages of relative robusticity in entheseal bony projections; see [10,39,50,54]). This question is also relevant to VERA 2.0, which automatically identifies and quantifies subtle projections within enthesis-bearing bone structures, even though it uses an initial BSR that includes a much larger portion of the bone (e.g., Fig 1 and example delineation video at protocols.io). According to all analyses and trials conducted in this validation study (see Figs 1–6 and their example illustrations), which involved numerous individuals and diverse species, VERA 2.0’s automatically identified surface irregularities were always predominantly occurring within the enthesis-bearing bone structure under analysis (e.g., within the body of the targeted tubercles or ridges).

Although the analyses of subsections b and c support VERA 2.0’s ability to distinguish among different activity groups, similar to previous experimental work with VERA 1.0 (see animal laboratory studies by [14–17]), the precise anatomical basis for its calculated surface changes remains an open question. Future anatomical and histological research is essential to confirm whether (or not) these automatically identified surface irregularities on bone structures correspond to the exact footprint of the soft tissues responsible for these deformations. While this issue falls outside the scope of the present 3D morphometric study, which explicitly focuses on introducing an efficient new tool for quantifying bone surface changes that are already routinely used in bioarchaeological contexts, I have included here a pilot anatomical dissection only as a preliminary step toward addressing this question in the future. This preliminary case-study only aims to lay a first foundation for future investigations into the relationship between 3D surface irregularities and soft tissue attachment, which could potentially enhance the multidisciplinary value of VERA 2.0 beyond bioarchaeological inferences of activity.

This case-study involved the dissection and virtual examination of the right OP’s musculotendinous unit and its insertion on the thumb metacarpal. It was performed using a fresh cadaveric hand from an adult male body donor under 60 years of age, with no manual pathological conditions. The dissection, carried out at the Medical School of Athens, Greece (see Acknowledgements), is illustrated in Fig 7. The hand was first positioned with the thumb and index fingertips brought in contact, to highlight the position of the thumb metacarpal and the thenar musculature involved in thumb opposition (Fig 7A). Once the skin and fat layers were removed, the superficial thenar muscles were exposed (Fig 7B). Further dissection (Fig 7C) revealed OP’s musculotendinous unit, which was manually pulled upwards, identifying four attachment points on the bone (labeled in the panel). These were: (1) a broad insertion site spanning from the metacarpal head to the bone’s midshaft (lateropalmar aspect), (2) a second and smaller insertion site in the proximal metacarpal (lateropalmar aspect), (3) an origin site on the trapezium’s tubercle, and (4) an additional large origin area on the flexor retinaculum. After the removal of the OP musculotendinous unit, the underlying bone and fibrocartilaginous layers were exposed (Fig 7D). The metacarpal was then surgically extracted, and a high-resolution 3D surface model was created, including surface texture and color (Fig 7E). This model highlighted the presence of a thick fibrocartilaginous layer (shown in brown color) within insertion areas 1 and 2 (outlined in the panel), which tended to be thicker in insertion area 1 (where both the OP muscle and joint capsular ligament used to attach). Before removal, the attachment of OP covered the predominant part of this area, while the capsular ligament only attached along the distal margins of the fibrocartilaginous layer, near the metacarpal head’s articular surface (as it can be partly observed in Fig 7D). Unfortunately, after the removal of these two soft tissues (see Fig 7E), visually distinguishing between their two exact attachment sites on the continuous and uniform surface of the fibrocartilaginous layer was impossible. Subsequently, micro-computed tomography was used to scan the metacarpal at a resolution of 40 microns. Through volume rendering and segmentation processes (using the software Avizo; Thermo Fisher Scientific, Massachusetts, United States), the fibrocartilaginous layers were virtually removed to reveal the underlying dry metacarpal bone (Fig 7F). Interestingly, before using VERA’s advanced 3D imaging filters, the resulting bone model only exhibited a small bone tubercle beneath the soft tissue layer, and only within the distal half of insertion area 1 (see Fig 7F), revealing that the actual soft tissue attachment region of OP and its supporting fibrocartilaginous layer were substantially larger than the visible underlying bone projection. VERA 2.0 was then applied to the OP’s enthesis, automatically detecting the 3D entheseal changes marked in blue (Fig 7G). These changes largely corresponded to both OP insertion regions (1 and 2) and their associated fibrocartilaginous layers, demonstrating VERA 2.0’s ability to detect subtle surface irregularities not clearly visible to the naked eye (even in high-resolution 3D models). To further highlight this observation, the 3D entheseal changes were virtually superimposed on the original 3D surface model (using Avizo tools), which still included the fibrocartilaginous layer (Fig 7H). This confirmed that the entheseal irregularities detected by VERA 2.0 (within a broadly defined BSR) largely corresponded to irregularities within the actual footprint of the attaching soft tissues. Overall, this analysis demonstrates that VERA 2.0 has the capacity to accurately identify subtle entheseal changes beneath the soft tissue attachment footprints, even when the defined BSR includes a much larger area and regardless of the bone’s visible gross morphology.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Dissection and virtual analysis of the opponens pollicis (OP) muscle’s attachment on the thumb metacarpal (all photographs and pictures of this figure are unpublished and were produced and edited by the author; see Acknowledgements).

(A) Cadaver hand placed in thumb opposition, (B) Exposure of superficial thenar muscles, (C) OP musculotendinous unit with attachment origin and insertion points labeled (1-4), (D) Thumb metacarpal bone and overlaying fibrocartilaginous layers exposed, (E) 3D surface model of the metacarpal’s lateropalmar aspect (distal is up), showing the brown-looking thick fibrocartilage layers within OP’s insertion regions (note: the two approximate main insertion areas are outlined), (F) Micro-computed tomography 3D model of the bone after virtually removing the soft tissue layers, (G) VERA 2.0 detection of entheseal changes (in blue), (H) Superimposed VERA 2.0 changes on the bone prior to virtual segmentation (i.e., soft tissue still present). More details are provided in the text. Scale is not respected across panels, but the metacarpal’s maximum length is 45 mm.

https://doi.org/10.1371/journal.pone.0321479.g007

As highlighted above, despite these impressive preliminary observations, I consider this subsection's anatomical case-study (dissection) only as the first step in this new line of anatomical and histological research, and I encourage future interdisciplinary studies to study additional entheses of the human body, after seeking and confirming access to larger fresh cadaveric samples of body donors with documented activities and non-pathological musculotendinous units.

Additionally, it is crucial to highlight again that VERA 2.0 could not distinguish between different soft tissue attachments within the same BSR, as exemplified by the OP muscle and joint capsular ligament in this dissection. This limitation was not surprising, given that there was no clear visible distinction between the borders of these two structures on the thick fibrocartilaginous layer (Fig 7E) or the underlying bone surface (Fig 7F), even in this fresh and healthy cadaveric hand with perfectly preserved soft tissue. Despite this limitation (which is likely inherent to all osteological analyses), VERA 2.0’s potential ability to identify bone surface changes related to soft tissue attachments remains highly relevant for inferring lifetime physical activity. In cases where different tissues attach in the same BSR region, observers should recognize and discuss that the detected entheseal changes may reflect the combined biomechanical influences (stress) of all structures involved. This does not necessarily impede the nature or value of functional interpretations (as previously discussed [19,12,21,23]). For example, humanlike thumb opposition routinely involves both OP contraction and strong flexion of the first metacarpophalangeal joint, to which the capsular ligament directly contributes [55–58], suggesting that the stress experienced by these two neighboring structures is hardly uncorrelated (see relevant discussion in [21]). Therefore, while VERA 2.0 (or any other anthropological approach to date) may not be able to automatically isolate individual attachments within a given bone BSR, its results will usually still be functionally meaningful (e.g., Figs 4–6), as neighboring or overlapping soft tissues often tend to closely interact during movement. This principle was consistently demonstrated in previous studies using VERA 1.0, which frequently analyzed enthesis-bearing structures corresponding to more than one muscle, such as for example the lateral basal tubercle of the human thumb’s proximal phalanx, which accommodates the insertion sites of the two synergist thenar muscles abductor pollicis brevis and flexor pollicis brevis [3,12,13,18,19,21,23–26].