The identification of human skeletal remains by frontal sinus morphology using three dimensional data obtained from computed tomography (CT scans)

In 1993, a court case in the United States of America, Daubert vs Merrell Dow Pharmaceuticals rewrote the rules by which scientific evidence was accepted as part of criminal proceedings. Many traditional forensic techniques, such as DNA, fingerprints, handwriting and bite-mark analysis have been intensely scrutinized since the Daubert case. For the most part, the problem lies with the limited information available concerning the degree of error inherent in each technique. This study was designed to address some of the issues raised by the Daubert standards in relation to the use of frontal sinus imaging for forensic identification. Specifically, it tests the reliability and discriminating power of data obtained from CT scans of the frontal sinus. This is achieved in three stages; firstly by proposing a method to compare frontal sinuses by using simple mathematical derivatives of the frontal sinus form, rather than attempting to accurately capture an exact morphology. These simple data were used to create a classical multi-level hierarchical comparison protocol, in which the fastest and least complex sinus derivatives are used to distinguish cases first, and progresses to more labour intensive and complex data analysis only when simpler measures fail. Comparisons commenced with the two sinuses (left and right) together, and proceeded to the contributions of the left and right sinuses independently, before examining information section-by-section. This method was able successfully to discriminate all of the individuals within the sample population of 56 individuals and allowed for cases where only unilateral information was available.

Secondly, by determining the degree of error that lies within components of the image analysis process and, hence, establishes the reliability of each. The sources of error likely to have an impact on the ability to distinguish individuals included; imaging section thickness, degree of section tilt, the condition of the remains i.e. fleshed vs skeletal, machine and display biases, re-measurement and inter-measurer error and automated compared with manual sinus feature detection. It was found that sources of error, except those arising from judgement and decision-making, were small in relation to differences between individuals. Systematic variations, such as differences between the sexes and changes with age, were also explored and found to be of little help in aiding the identification.

The third phase of this study tested the application of the protocol to a real life data set comprising a database of frontal sinus information from 56 individuals, much like those used for fingerprint identification. The protocol of sinus comparison developed in the first stage of the study successfully matched sinus data from a number of “unknown” test cases from the dataset to their respective cases in the database excluding all others. No matches were identified when “unknown” test cases were derived from sources outside the original database i.e. no “false positive” matches were obtained. Measurements taken from fresh-frozen cadaveric specimens allowed successful and exclusive matching with data derived from skeletonised remains and from ante-mortem clinical scans. This study establishes the use of frontal sinus data from CT scans as a valid and reliable method of identification.