A real-coded evolutionary algorithm-based registration approach for forensic identification using the radiographic comparison of frontal sinuses ´ Oscar G´ omez *†‡ , Pablo Mesejo *†‡ , ´ Oscar Ib´ a˜ nez †‡ , Andrea Valsecchi †‡ and ´ Oscar Cord´ on *† * Dept. of Computer Science and Artificial Intelligence, University of Granada, Spain. † Andalusian Research Institute DaSCI, University of Granada, Spain. ‡ Panacea Cooperative Research S. Coop., Ponferrada, Spain. Email: ogomez@decsai.ugr.es, pmesejo@decsai.ugr.es, oscar.ibanez@panacea-coop.com, valsecchi.andrea@panacea-coop.com, ocordon@decsai.ugr.es Abstract—Comparative radiography is the forensic anthropol- ogy technique in which ante-mortem (AM) and post-mortem (PM) radiographic materials (e.g., X-ray images or CTs) are compared in order to determine the identity of a deceased human being. One of the most commonly used anatomical structures in comparative radiography are the frontal sinuses. The frontal sinuses are osseous cavities located in the skull, which are used in forensic identification tasks due to their singularity and high identification power. In order to automate the comparison of frontal sinuses in AM and PM materials, it is necessary to perform the registration of these materials (i.e., it is necessary to carry out the alignment of these anatomical regions). However, the manual alignment of these structures is a time-consuming and subjective process. In order to tackle this problem, this paper presents an automatic frontal sinuses registration method in comparative radiography using real-coded evolutionary algorithms (RCEAs). The task is formulated as a 2D-3D image registration problem using a 9 Degrees of Freedom perspective transformation model; two RCEAs (DE and MVMO- SH) are compared in the minimization of the registration cost function, and the best of them (MVMO-SH) is applied to an identification scenario including 50 X-ray images and 50 CTs. The results obtained show that the proposed automatic identification system is able to filter more than 80% of the sample. Index Terms—Biomedical Image Registration, Forensic Identi- fication, Comparative Radiography, Frontal Sinuses, Real-Coded Evolutionary Algorithms, Mean Variance Mapping Optimization, Differential Evolution I. I NTRODUCTION Comparative radiography (CR) [1] is a forensic identifica- tion technique based on the comparison of skeletal structures in ante-mortem (AM) and post-mortem (PM) radiographs. Since the discovery of X-rays by Roentgen in 1895 [2], forensic experts have made use of radiographic images as evidence in their endeavour (e.g. bullet analysis [3], age estimation [4], and forensic identification [5]). During the first decades of the twentieth century, the use of X-rays as a method of positive identification gradually consolidated in scientific literature. In fact, in 1949, CR techniques played a crucial role in the identification of people involved in the Noronic ship’s disaster, proving their importance for identification and being included in many mass disaster identification protocols [6]. Nowadays, CR is still employed in many forensic identi- fication scenarios. For instance, the Michigan State University Forensic Anthropology Laboratory (MSUFAL) performed 193 identifications using this approach between 2002 and 2015 [7]. Several bones and cavities have been reported as useful for candidate short-listing or positive identification based on their individuality and uniqueness [8]. In particular, frontal sinuses (see Fig. 1) are widely recognized as a useful and reliable method of identification [9], fulfilling the Daubert criteria [10] 1 . Frontal sinuses are only absent in 4% of the population and are maintained unchanged during the rest of the life [12]. Although, rarely, some external factors such as traumatisms can change slightly their morphology, frontal sinuses are con- sidered as a skeleton fingerprint. Their utilization for CR-based identification was first reported in 1926 by comparing their morphology in AM and PM radiographs [13]. Nowadays, CR identification based on frontal sinuses is widely accepted by the forensic community, and many works have reported their utility via image comparison to establish positive identification [14]–[16]. CR techniques have lower cost and time requirements in comparison to DNA analysis, which are crucial factors in mass disaster victim identification scenarios. However, the applica- tion of CR requires the superimposition of the AM and PM data for their visual comparison by producing PM radiographs simulating the AM ones in scope and projection. This is a time-consuming trial-and-error process, that relies completely on the skills and experience of the analyst. Furthermore, the utility of the method is reduced because of the errors related to analysts’ fatigue and subjectivity. There is thus a need to automate CR-based identification methods. The automation of the CR’s superimposition process is complex and computationally expensive (see Section II for further details). This is due to several reasons, such as the unknown set-up of the AM radiograph, the fact that image 1 The Daubert criteria [11] determine whether evidence is admissible in a court of law. An identification method fulfills the Daubert criteria when: (1) it is testable and peer reviewed; (2) it possesses known potential error rates; and (3) it is accepted by the forensic community. 978-1-7281-6929-3/20/$31.00 ©2020 IEEE