Advances in Osteometric Sorting: Utilizing Diaphyseal CSG Properties for Lower Limb Skeletal Pair‐Matching

Abstract

Pair‐matching of bilateral elements is a major component of resolving commingled remains both in forensic and bioarchaeological contexts. This study presents a new method of osteometric pair‐matching of the lower limbs which relies on 3D digital models of the femur and tibia bones. The proposed method, which is accompanied by a freely available open‐source implementation, automatically computes a number of osteometric variables including cross‐sectional geometric properties from an assemblage of left and right bone antimeres and calculates probabilistically the appropriate matching pairs as well as single elements, whose bones antimere is not present in the given assemblage. The method has been extensively tested on a skeletal sample comprising 396 femurs and 422 tibias from the Athens collection. Our results in testing commingled assemblages with no disparity show that the method’s sensitivity is 1 for sorting femurs and 0.997 for sorting tibias, whereas in assemblages with moderate disparity the sensitivity is 0.999 and 0.992 respectively. Our results further indicate that sensitivity is unaffected by the size of the commingled assemblage although the percentage of identified true matching pairs drops as the number of commingled elements increases. This means that all identified antimeres matched to an individual are still very accurately sorted despite not every individual being identified in very large assemblages. The proposed method can facilitate the sorting process of commingled remains both accurately and efficiently, while leaving a very small percentage of unsorted elements that may require further techniques for further individualization.

Next‐generation osteometric sorting: Using 3D shape, elliptical Fourier analysis, and Hausdorff distance to optimize osteological pair‐matching

“…Typically a 10% error rate, for wrongly excluding ground-truth anatomical pairs is accepted as standard [1,13,16,20]. High rates of failure to exclude skeletal elements that are not true pairs are a widely recognized limitation [7,16,18,24]. Subsequently, complex statistical modeling has been pursued as one potential solution to boost exclusion power.…”

“…Recently, Bertsatos and Chovalopoulou [7] described a new approach using the long-bone-diaphyseal-CSG-Toolkit [36], which is open-source software, and tested it on femora and tibiae of multiple samples comprised of up to 120 individuals (drawn from a larger sample of N = 230 individuals). The CSG method uses 46 variables defining maximum distance, cross-sectional areas, cross-sectional perimeters, robusticity (2nd moment of area for each cross-section), and shape related to diaphyseal bending (dihedral angles) [7]. Averaged true positive rates over 30 iterations of repeated sample draws, with up to 100 true pairs in the test samples, were reportedly 0.976-1.0 and true negative rates were in the vicinity of 0.985 [7].…”

“…The CSG method uses 46 variables defining maximum distance, cross-sectional areas, cross-sectional perimeters, robusticity (2nd moment of area for each cross-section), and shape related to diaphyseal bending (dihedral angles) [7]. Averaged true positive rates over 30 iterations of repeated sample draws, with up to 100 true pairs in the test samples, were reportedly 0.976-1.0 and true negative rates were in the vicinity of 0.985 [7].…”

“…Two broad approaches are pertinent to the unmingling of skeletons, namely pair-matching of bilaterally paired bones from the same body region and the association of other different bones often from different body regions. When large numbers of individuals comprise the commingled assemblage, manual sorting processes can be laborious (even if just pair-matching the major long bones), requiring many tedious and repetitive measurements/calculations [3][4][5][6][7].…”

“…The single largest limitation to morphology-based unmingling methods is their lower sensitivity and generally higher error rates in larger more morphologically homogeneous samples. This limitation has generally seen morphologybased osteometric sorting methods regarded for their exclusionary capabilities, rather than for direct or individual-specific assignment of paired elements [7,16,18].…”