Dissertation committee:
Ε. Βαλάκος, Καθηγητής, Πανεπιστήμιο Αθηνών (Επιβλέπων)
Χ. Παπαγεωργοπούλου, Αν. Καθηγήτρια, Πανεπιστήμιο Θράκης
Ι. Λυριτζής, Καθηγητής, Πανεπιστήμιο Αιγαίου
Ε. Νικήτα, Αν. Καθηγήτρια, Ινστιτούτο Κύπρου
Α. Κακάμπουρα, Καθηγήτρια, Πανεπιστήμιο Αθηνών
Π. Παφίλης, Αν. Καθηγητής, Πανεπιστήμιο Αθηνών
Δ. Χατζηνικολάου, Αν. Καθηγητής, Πανεπιστήμιο Αθηνών
Summary:
The purpose of this doctoral study was to expand our knowledge on the diversity of the human skeleton of a modern Greek population and to develop new methodologies and tools for extracting biological parameters, such as sex and age-at-death, from human skeletal remains. The whole research was based on the modern skeletal reference collection, housed in the Physical Anthropology laboratory of the Department of Animal and Human Physiology of the Department of Biology, which is the most complete skeletal reference collection in Greece and is considered a representative sample of the Greek population of the 20th century. A major goal has been the digital documentation of a subset of the available skeletal sample, which an array of individual research questions focused upon. The digital documentation, based on three-dimensional photogrammetry, covers the skull, the mandible, the os coxae, the femurs, the tibiae, and the humeri from all adult individuals of the skeletal collection, where available. The completion of the digital documentation yielded 2,117 high-precision three-dimensional mesh models with high-resolution photorealistic texture. These were used as a sample in numerous osteometric analyses, many of which cannot be applied to dry bones with classical osteometric methods. Furthermore, they are a consignment, as a digital library, for future research while allowing for optimal storage and maintenance of the modern skeletal collection, to which physical access is no longer necessary.
A series of studies were conducted focusing on the biological parameters of sex and age-at-death within the doctorate. In this light, the expression of cranial sexual dimorphism was studied and employing the optimal metric characteristics, population-specific discriminant function equations for estimation of sex have been formulated, which gave an identification accuracy of over 95%. A similar study was conducted for the mandible, which can facilitate sex estimation of an unknown individual, despite being less sexually dimorphism compared to the skull. This study also revealed that the expression of the sexual dimorphism of the mandible is highly population specific.
In addition, the morphological variation of the fovea capitis femoris with respect to sex and age-at-death was investigated, and although it may not to be suitable for estimating sex or age-at-death in the forensic context, it was shown that the size is directly related to sex (systematically larger in males) while its shape with age-at-death (more irregular perimeter with increasing age). Furthermore, the secular change in stature of the Greek population for the reference period (1879-1965) was studied utilizing the maximum femoral length. According to the results, the stature increases over the total range of the examined period as expected. However, male individuals were either more vulnerable or received greater environmental stress from the socio-economic conditions associated with the wars of the first half of the 20th century as opposed to females.
During the development of the digital skeletal collection, a study was performed on the accuracy and precision in obtaining osteometric data from virtual (digital) skeletal remains using skull models. This study revealed that the use of three-dimensional models with photorealistic texture created with photogrammetry not only are reasonable substitutes for dry bones but in many cases allow for better measurement accuracy compared to manual osteometric instruments. Further exploiting the 3D cranial models, an evaluation of the reliability of 3D-ID software in sex and ancestry estimation was performed. The results showed a very low percentage of correct classification in narrowly defined groups, outlining certain weaknesses of the software but also the limitations of the adopted method (3DGM). However, the performance of 3D-ID software increases considerably when it comes to large population groups (eg Africans, Asians, Europeans, etc.).
In order to make the best use out of the three-dimensional models, a novel method for automatically computing the long bones’ diaphyseal cross-sectional geometric (CSG) properties was developed. This method, called the "long-bone-diaphyseal-CSG-Toolkit", was implemented as a free open source software aiming to replace the costly and time-consuming latex cast method (LCM) while at the same time it eliminates the measurement error resulting from the implementation of the latter. The CSG-Toolkit was further used to develop an automated method for estimating sex from the lower and upper limbs (femur, tibia, humerus) of the Greek population, which resulted in a classification accuracy reaching 97.3% for the humerus. Additionally, the long bones’ diaphyseal cross-sectional geometric properties along with the categorical data from the entheseal changes of the femur and tibia, being evaluated as skeletal activity markers, were used to study the impact of physical activity on specific pelvic age indicators, which are widely used to estimate age-at-death.
The latest research work of this doctorate concerns the development of an innovative method for estimating sex from virtual skulls in an automatic way, which eliminates any observer error. This new method combines the accuracy and repeatability of metric characteristics with the diagnostic capability of macroscopic morphological analysis to identify subtle differences between males and females by using machine learning techniques and new morphometric features inspired by existing computer vision methods. The new method, called "skullanalyzer", is also implemented as free open source software and was evaluated using two European population samples. According to the results, "skullanalyzer" yields accuracy in excess of 95% and is less prone to population specificity, which demonstrates a clear advantage over existing methods.
Keywords:
Physical and Forensic Anthropology, modern Greek population, digital documentation of skeletal remains, 3D photogrammetry, 3DGM, sex and age-at-death estimation, machine learning, free open-source software (libre)
