Abstract
Forensic facial approximations on to skulls are used for victim identity when there is a lack of personal items or dental records, or the face has decomposed past recognition. Yet identification success rate is not much above chance, and current methods are criticised for their subjectivity. All approximation methods, manual or computerised, include the use of Facial Soft Tissue Depth (FSTD) Tables that contain numerical values to guide tissue depth onto a skull. It is these depth values that this thesis focuses on.
A literature review examined published FSTD Tables from 1898 to 2013. It focused on measurement methods used to obtain FSTD values from living or deceased subjects, variables used to group subjects, and statistical accuracy of those variables. It was evident there is a lack of clarity regarding significant influences on FSTDs and that Table values have unexplained biases. It questioned whether using averaged data for an individual skull significantly decreases predictive accuracy, and whether the most common categorisation heirachy of FSTDs to ancestry, sex and then age variables is justifiable.
Quantitative influence of ancestry, Body Mass Index (BMI), position and age variables on FSTDs were investigated by two sonographic studies. Results from both studies showed that BMI was the most important variable to influence FSTDs. Ancestry FSTD differences between Chinese and New Zealand European cohorts were evident at the buccal (lateral mid cheek and posterior mandibular) region; where Chinese women showed significantly greater mean values, 3.5 mm, compared with NZ European (same BMI Classes). But for all other regions of the face, ancestry showed no significant difference. Supine postural position (gravitational influence) on depths was variable between individuals, and age difference between young and middle aged adults was not significant. These results affirmed the validity of questions asked in the literature review.
Chapters Two, Three and Four results stimulated inquiry to consider an alternative categorisation of skull ancestry to the common Caucasoid, Mongoloid or Negroid; acknowledged by bioarcheologists as problematic. Chapter Five investigates clustering FSTDs with similar depth and variance values to specified bone regions of the face, using Chapter Three data. Dendogram results showed ‘regional bone morphology’ could be an alternative option to categorising an entire skull to one ancestry. This conclusion stimulated development of novel protocol that measured both craniometric (bone) and skin landmarks from virtual 3D images, acquired from Cone Beam CT scans of living, healthy adults. Chapter Six details how 54 landmarks per subject, described as x,y,z co-ordinates, were acquired from 79 scans.
Chapter Seven describes transferance of 47 landmark data sets, acquired in Chapter Six, to a statistical predictive model, then FSTD predictive strength evaluated in one-to-one regressive predictive equations. Results show promise depth prediction, even without data separated according to BMI, age or sex variables. The primary advantage is that prediction can be tailored to each individual skull. The skull bone surface x,y,z values can be compared, using a priori knowledge with all other bone and skin x,y,z data sets already inputted; and the most likely soft tissue depths for that particular skull would then be calculated and presented. This model, if further refined, could mitigate computerised approximation reliance on externally sourced, averaged FSTD values; with accompanying biases and uncertain error.
Conclusions from this thesis are that soft tissue depth prediction for an individual skull could be increased in accuracy if there are changes to the method. This includes categorising FSTD tables according to correct, heirarchical ranking of variable influences; always quantifying BMI; describing the morphology of a skull according to separate facial regions, rather than grouping an entire skull to one ancestry group; and then linking each separate skull region with a cluster of FSTD values that have been obtained from subjects with a similar regional facial bone morphology. Finally, it is suggested that regressive prediction equations that take note of all, and each, values within a data set be used to predict a set of soft tissue depth values for an individual. Such changes should increase predictive accuracy of soft tissue depth, with the intention to achieve an increased rate of success for a forensic identification.