Conventional methods of body fluid identification use labor-intensive, technologically diverse techniques that are performed in a series, not parallel, manner and are costly in terms of time and sample. Furthermore, for some frequently encountered body fluids, such as saliva or vaginal secretions, no confirmatory technique exists. Terminally differentiated cells, such as blood lymphocytes or epithelial cells lining the oral cavity, have a unique pattern of gene expression, which is evinced by the presence and relative abundance of specific mRNA species. If the type and abundance of mRNAs can be determined in a stain or tissue sample recovered at the crime scene, it would be possible to definitively identify the tissue or body fluid in question. Advantages of an mRNA-based approach, compared to conventional biochemical analysis, include greater specificity, simultaneous and semi-automated analysis though a common assay format, improved timeliness, decreased sample consumption and compatibility with DNA extraction methodologies. In this report, we demonstrate that RNA is stable in biological stains and can be recovered in sufficient quantity and quality for analysis using reverse transcriptasepolymerase chain reaction assay (RT-PCR). We have identified sets of candidate tissuespecific genes for body fluids and tissues of forensic interest, namely blood, saliva, semen, vaginal secretions, menstrual blood, urine, skin, muscle, adipose, and brain. We also report the identification of a new housekeeping gene for use in mRNA based assays. Select body fluid-specific genes have been incorporated into multiplex PCR and real-time PCR assays. These assays allow for the positive identification of blood, saliva, semen,vaginal secretions, and/or menstrual blood in a stain. The final task of this work was the molecular characterization of mRNA degradation patterns in biological stains, which not only has fundamental importance in possibly revealing mRNA degradation pathways in dried biological stains, but may ultimately lead to better assay design strategies for mRNA markers for forensic use. An mRNA-based approach described in this report could allow the facile identification of the tissue components present in a body fluid stain and could conceivably supplant the battery of serological and biochemical tests currently employed in the forensic serology laboratory.

Pour analyser les images en tomodensitométrie, une méthode stœchiométrique est gé- néralement utilisée. Une courbe relie les unités Hounsfield d’une image à la densité électronique du milieu. La tomodensitométrie à double énergie permet d’obtenir des informations supplémentaires sur ces images. Une méthode stœchiométrique a été dé- veloppée pour permettre de déterminer les valeurs de densité électronique et de numéro atomique effectif à partir d’une paire d’images d’un tomodensitomètre à double énergie. Le but de cette recherche est de développer une nouvelle méthode d’identification de tissus en utilisant ces paramètres extraits en tomodensitométrie à double énergie. Cette nouvelle méthode est comparée avec la méthode standard de tomodensitométrie à simple énergie. Par ailleurs, l’impact dosimétrique de bien identifier un tissu est déterminé. Des simulations Monte Carlo permettent d’utiliser des fantômes numériques dont tous les paramètres sont connus. Les différents fantômes utilisés permettent d’étalonner les méthodes stœchiométriques, de comparer la polyvalence et la robustesse des méthodes d’identification de tissus double énergie et simple énergie, ainsi que de comparer les distributions de dose dans des fantômes uniformes de mêmes densités, mais de compo- sitions différentes. La méthode utilisant la tomodensitométrie à double énergie fournit des valeurs de densi- tés électroniques plus exactes, quelles que soient les conditions étudiées. Cette méthode s’avère également plus robuste aux variations de densité des tissus. L’impact dosimé- trique d’une bonne identification de tissus devient important pour des traitements aux énergies plus faibles, donc aux énergies d’imagerie et de curiethérapie. / A stoichiometric method is usually used to analyze computed tomography images. A curve links the Hounsfield units on the images to the electron density in a given me- dium. Dual-energy computed tomography gives additional information on a scan. A stoi- chiometric method was developed to acquire both electron density and effective atomic number from a pair of images. The aim of this research is to develop a new method to identify tissues using the parame- ters extracted from dual-energy computed tomography. This new method is compared to the standard single-energy computed tomography segmentation method. Furthermore, the effect of correctly assigning tissues on dose distribution is studied. Monte Carlo simulations allow the use of perfectly known numerical phantoms. Dif- ferent phantoms allowed the calibration of the stoichiometric methods, the comparison of the versatility and the robustness of the dual-energy and the single-energy methods, and the comparison of dose distribution in phantoms of same densities, but of different compositions. The dual-energy identification method gives more accurate values of electron density in any studied condition. This method is also more robust to tissues of variable density. The dosimetric impact of an accurate identification becomes more important for treatments using lower energy photons, such as imaging energies and brachytherapy.