人工神经网络的法医学应用

人工神经网络(ANN)是对人脑结构及功能的数学模拟系统,具有强大的非线性拟合能力和自学习、容噪声等特点,是一种全新的数据处理模式。近年来,ANN在各学科得到了广泛研究和应用。本文从应用角度概要介绍ANN及国外法医学应用现状,旨在推动国内法医学研究与开发。     

Optimization of the separation of organic explosives by capillary electrophoresis with artificial neural networks

The separation of 12 explosives by capillary electrophoresis was optimized with the aid of artificial neural networks (ANNs). The selectivity of the separation was manipulated by varying the concentration of sodium dodecyl sulfate (SDS) and the pH of the electrolyte, while maintaining the buffer concentration at 10 mM borate. The concentration of SDS and the electrolyte pH were used as input variables and the mobility of the explosives were used as output variables for the ANN. In total, eight experiments were performed based on a factorial design to train a variety of artificial neural network architectures. A further three experiments were required to train ANN architectures to adequately model the experimental space. A product resolution response surface was constructed based on the predicted mobilities of the best performing ANN. This response surface pointed to two optima; pH 9.0-9.1 and 60-65 mM SDS, and pH 8.4-8.6 and 50-60 mM SDS. Separation of all 12 explosives was achieved at the second optimum. The separation was further improved by changing the capillary to an extended cell detection window and reducing the diameter of the capillary from 75 microm to 50 microm. This provided a more efficient separation without compromising detection sensitivity.     

Homicide-suicide in the change of time

Homicide-suicide is a rare event compared to suicide and homicide alone. A comparison of the statistical figures of two studies for Hamburg (1950-1961 and 1990-2005) shows a significant development within the last 40 years: the total number of homicide-suicides committed by women has decreased considerably, probably due to a better medical and psychiatric care and greater economic independence, whereas the number of male perpetrators has remained constant. However, these perpetrators are younger now. As far as the methods are concerned, shooting is now the most common means of killing instead of poisoning by coal gas in former days.     

Comparative study of Greulich and Pyle Atlas and Maturos 4.0 program for age estimation in a Portuguese sample

Age estimation of living individuals has become one of the big issues of forensic anthropology. The increase of children and adolescents with no valid proof of their chronological age is a legal concern to many countries, especially in situations of illegal immigration, sub adult delinquency and juvenile work. For this purpose, the use of radiological methods for evaluation of skeletal maturation is particularly useful. In this study we compare the two of the most common methods applied in age estimation by hand/wrist radiographs: the Radiographic Atlas of Skeletal Development of the Hand and Wrist made by Greulich and Pyle (GP), 1959,and the Sempé method developed for computer--Maturos 4.0 (MT) program. These methods were applied to a sample of 230 radiographs of the hand and wrist from Portuguese children and adolescents of known sex age and age, aged between 12 and 20 years, who performed medical examination at the University of Coimbra Hospitals during 2005. The methods achieved different performances, depending on the age group. Between 12 and 15 years the bone age with the MT program is closer to the chronological age, whereas in older ages the GP Atlas method is more trustworthy. At the ages with legal consequences in Portugal (16 and 18 years) the GP Atlas method is most accurate, namely to decide whether an individual is younger or older than 18 years. Around 16 years old, although there are doubts in relation to the accuracy of both methods, GP Atlas seems to perform better.     

Neural networks in forensic science.

Neural networks were developed to study and mimic the functioning of the human brain. Humans are good at pattern recognition; the question is how good neural networks are at it, particularly with problems of forensic science interest. Simulation experiments with a type of neural network known as a Hopfield net indicate that it may have value for the storage of toolmark patterns (including bullet striation patterns) and for the subsequent retrieval of the matching pattern using another mark by the same tool for input. Another type of neural network, the back-propagation network (BPN), is useful for applications similar to those for which standard statistical methods of pattern classification can be used. This would be an appropriate approach to the matching of general component patterns, such as gas chromatograms of gasoline, or pyrolysis patterns from materials of forensic science interest, such as paint. The BPN may provide better results than statistical methods, but it is currently necessary to try both to determine which would be best for any given situation.     

Violence and trauma: the past 20 and next 10 years

The issues of violence and trauma over the past 20 years have generated a significant amount of research. However, the development of each field has remained, for the most part, exclusive from the other. Therefore, the goal of this article is to address this shortcoming within the literature by responding to three important questions: (a) What is the most important thing we have learned about violence and trauma in the past 20 years? (b) What is the most important thing we need to learn in the next 10 years? and (c) What is the most promising methodological innovation in the past 20 years for the study or treatment of trauma or interpersonal violence? This article concludes that by increasing the discourse between the two fields, both research fields may be better able to develop conceptually, methodologically, and empirically.     

The classification and discrimination of glass fragments using non destructive energy dispersive X-ray microfluorescence

Frequency of analytical characteristics is best estimated on glass recovered at random. However, as such data were not available to us, we decided to use control windows for this estimation. In order to use such a database, one has to establish that the recovered fragment comes from a window. Therefore, elemental analysis was used both for classification and discrimination of glass fragments. Several articles have been published on the subject, but most methods alter the glass sample. The use of non destructive energy dispersive X-ray microfluorescence (microXRF) for the analysis of small glass fragments has been evaluated in this context. The refractive index (RI) has also been measured in order to evaluate the complementarity of techniques. Classification of fragments has been achieved using Fisher's linear discriminant analysis (LDA) and neural networks (NN). Discrimination was based on Hotelling's T2 test. Only pairs that were not differentiated by RI followed by the Welch test were studied. The results show that neural network and linear discriminant analysis using qualitative and semi-quantitative data establishes a classification of glass specimens with a high degree of reliability. For discrimination, 119 windows collected from crime scene were compared: using RI it was possible to distinguish 6892 pairs. Out of 129 remaining pairs, 112 were distinguished by microXRF.     

Protecting democracy: a parsimonious,dynamic and heuristic model of controlling crimes by the powerful*

Crimes of the powerful (i.e. corporations, elites, organized crime groups, states, white collar criminals, etc.) have numerous devastating effects. That is why, over the centuries, some philosophers and political theorists - and more recently some constitutional experts, and selected social scientists - have proposed and analyzed the best ways to deter, minimize, prevent, regulate, and hereafter control large powerful interests. Not all powerful actors, however, are equally dangerous nor does this threat remain constant. Moreover, not all methods of control work for all entities. In order to better understand this process, one must determine a number of things including: which powerful actor is currently the most dangerous/harmful, where controls do not exist and/or do not function well, and which controls work best and in which particular setting. To achieve this goal, scholarship in diverse fields need to be integrated, with specific reference to the challenges and problems that have arisen and the solutions that have been proposed. To this end, this article develops a parsimonious but dynamic and heuristic model, consisting of a number of interconnected factors, to depict the process of controlling crimes by the powerful.     

Evaluation of post-mortem estimated dental age versus real age: a retrospective 21-year survey

The aim of the study was to evaluate the reliability of methods used for forensic dental age estimation. We analysed all cases over the last 21 years (1984-2004) of unidentified bodies that were examined for identification purposes (including age assessment), and of which secure identification was subsequently achieved. In total, the study included 51 cases and 7 different methods had been used for dental age estimation, with the Bang/Ramm and the Gustafson/Johanson methods being the most frequently applied. The age estimates had usually been recorded as 10-year intervals. Factual ages at death were in the range of 6-76 years, with the largest concentration of cases being in the age interval of 25-55 years (34 cases). There was good agreement between estimated age interval and factual age at death in 37/51 (72%) of the cases. In eight cases the factual age at death deviated up to +/-5 years from the estimated age, and in six cases by more than 6 years. The average difference between factual age at death and estimated age was 4.5 years. The four subadults in the material were all correctly estimated within an age range of +/-3 years. Our study showed that forensic odontological age estimates are reliable. However, the implementation of the specific methods may need to be adjusted concerning age ranges. In the future we recommend to register anamnestic information and the different steps in the methods used. Clinical evaluation should contain more details about attrition, colour, number and presumed age of the restorations and periodontal status.     

Analysis of breath alcohol with a multisensor array: instrumental setup, characterization and evaluation

The purpose of this work was to evaluate if it was possible to measure the alcohol concentration in breath by a multisensor array, i.e. an electronic nose. The most important aspects were to clarify technical advantages and disadvantages and if the technique is at all suitable for forensic breath alcohol analysis. Even though the system set-up was far from optimal it is clear that it was possible to quantify breath alcohol with a best-case root mean square error = 16.8 mol ppm ethanol (= 0.037 mg/l). However, the method needs significant development especially of the sensor devices. The sensor array was composed of ten metal oxide silicon field effect transistors (MOSFET) with various catalytic gates and one infrared-based CO2 sensor. The system was evaluated by monitoring the breath ethanol concentration of five test persons after intake of alcohol. Gas chromatography was used in parallel to measure the actual breath alcohol concentration. Different data evaluation techniques applied were projection to latent structure (PLS) and artificial neural networks (ANN).     

Technology transfer between the U.S.A. and Europe

The premise is that technology gaps have an important impact on the economic life of nations and also have political consequences. Expressions of concern about the technology gap between Europe and the United States have become steadily less frequent in the recent years. The purpose of this paper is to find out whether some of the lessons that can be drawn from the European technology gap of the '60s contribute to the understanding of the present United States-European technology exchange controversies. This is accomplished by (1) reviewing the European arguments, (2) trying to find out how revelant they have proven to be after a few years, (3) investigating better ways to assess the impact of technological differences with reference to current United States arguments about technology export, and (4) attempting to derive some conclusions on policy implications of transferring technology. It was concluded that important policy decisions have been made and are still being made by technology importing countries with little analytical background on the cost and advantages of achieving a technological capacity. In addition, policy decisions by technology exporting countries are most likely to be made with little knowledge of the real phenomena involved.     

Is there recent progress in the estimation of the postmortem interval by means of thanatochemistry?

Numerous methods have been proposed in the last 60 years for the determination of the time since death by chemical means. Many of them were reviewed by Schleyer in his monograph on the determination of the time since death by means of thanatochemistry about 40 years ago and none of these early methods has gained any practical value since they do not meet the demands in practice (being precise, reliable, giving an immediate result). While the earlier studies were mainly carried out on blood and cerebrospinal fluid (CSF) since the late 60s most investigations have been performed on vitreous humor (VH). This is mainly due to the fact that vitreous humor is topographically isolated and well protected, and thus, autolytic changes proceed slower compared to blood and CSF. The most studied parameter in VH is potassium and even nowadays reports on the postmortem rise of vitreous potassium are published, proposing new analytical methods or statistical evaluations. Chemical parameters studied for the determination of the time since death have to be differentiated according to the underlying process (catabolism, metabolic processes, pure autolysis and diffusion, putrefactive changes). In the present paper, recent studies on thanatochemistry are discussed regarding the underlying process, the analytical methods (for instance H magnetic resonance spectroscopy (1H MR spectroscopy), immunohistochemistry), the studied fluid compartment, the statistical evaluation and the precision of death time estimation. The value of chemical methods for the determination of the time since death is up to now very limited. This is supported by the fact that field studies on the reliability and precision of death time estimation by chemical means are still scarce in the literature.     

Classification of premium and regular gasoline by gas chromatography/mass spectrometry,principal component analysis and artificial neural networks

Detection and correct classification of gasoline is important for both arson and fuel spill investigation. Principal component analysis (PCA) was used to classify premium and regular gasolines from gas chromatography-mass spectrometry (GC-MS) spectral data obtained from gasoline sold in Canada over one calendar year. Depending upon the dataset used for training and tests, around 80-93% of the samples were correctly classified as either premium or regular gasoline using the Mahalanobis distances calculated from the principal components scores. Only 48-62% of the samples were correctly classified when the premium and regular gasoline samples were divided further into their winter/summer sub-groups. Artificial neural networks (ANNs) were trained to recognise premium and regular gasolines from the same GC-MS data. The best-performing ANN correctly identified all samples as either a premium or regular grade. Approximately 97% of the premium and regular samples were correctly classified according to their winter or summer sub-group.     

Patella sex determination by 3D statistical shape models and nonlinear classifiers

Sex determination is one of the essential steps in personal identification of an individual from skeletal remains. Most elements of the skeleton have been subjected to discriminant function analysis for sex estimation, but little work has been done in terms of the patella. This paper proposes a new sex determination method from the patella using a novel automated feature extraction technique. A dataset of 228 patellae (95 females and 133 males) was amassed from the William M. Bass Donated Skeletal Collection from the University of Tennessee and was subjected to noninvasive high resolution computed tomography (CT). After the CT data were segmented, a set of features was automatically extracted, normalized, and ranked. The segmentation process with surface smoothing minimizes the noise from enthesophytes and ultimately allows our methods to distinguish variations in patellar morphology. These features include geometric features, moments, principal axes, and principal components. A feature vector of dimension 45 for each subject was then constructed. A set of statistical and supervised neural network classification methods were used to classify the sex of the patellar feature vectors. Nonlinear classifiers such as neural networks have been used in previous research to analyze several medical diagnosis problems, including quantitative tissue characterization and automated chromosome classification. In this paper, different classification methods were compared. Classification success ranged from 83.77% average classification rate using labels from a Fuzzy C-Means (FCM) clustering step, to 90.3% for linear discriminant classification (LDC). We obtained results of 96.02% and 93.51% training and testing classification rates, respectively, using feed-forward backpropagation neural networks (NN). These promising results using newly developed features and the application of nonlinear classifiers encourage the usage of these methods in forensic anthropology for identifying the sex of an individual from incomplete skeletons retaining at least one patella.     

The placement of the human eyeball and canthi in craniofacial identification

An accurate understanding of the spatial relationships between the deep and superficial structures of the head is essential for anthropological methods concerned with the comparison of faces to skulls (superimposition) or the prediction of faces from them (facial approximation). However, differences of opinion exist concerning: (i) the position of the eyeball in planes other than the anteroposterior plane and (ii) the canthi positions relative to the bony orbital margins. This study attempts to clarify the above relationships by dissection of a small sample of adult human cadavers (N = 4, mean age = 83 years, s = 12 years). The most notable finding was that the eyeballs were not centrally positioned within the orbits as the more recent craniofacial identification literature expounds. Rather, the eyeballs were consistently positioned closer to the orbital roof and lateral orbital wall (by 1-2 mm on average); a finding consistent with the earlier anatomical literature. While these estimation errors are small ipsilaterally, several factors make them meaningful: (i) the orbital region is heavily used for facial recognition; (ii) the width error is doubled because the eyes are bilateral structures; (iii) the eyes are sometimes used to predict/assess other soft tissue facial structures; and (iv) the net error in facial approximation rapidly accumulates with the subsequent prediction of each independent facial feature. While the small sample size of this study limits conclusive generalizations, the new data presented here nonetheless have immediate application to craniofacial identification practice because the results are evidence based. In contrast, metric data have never been published to support the use of the central positioning guideline. Clearly, this study warrants further quantification of the eyeball position in larger samples and preferably of younger individuals.     

Human Leukocyte Antigen alleles as an aid to STR in complex forensic DNA samples

Human biological samples with multiple contributors remain one of the most challenging aspects of DNA typing within a forensic science context. With the increasing sensitivity of commercially available kits allowing detection of low template DNA, complex mixtures are now a standard component of forensic DNA evidence. Over the years, various methods and techniques have been developed to try to resolve the issue of mixed profiles. However, forensic DNA analysis has relied on the same markers to generate DNA profiles for the past 30 years causing considerable challenges in the deconvolution of complex mixed samples. The future of resolving complicated DNA mixtures may rely on utilising markers that have been previously applied to gene typing of non-forensic relevance. With Massively Parallel Sequencing (MPS), techniques becoming more popular and accessible even epigenetic markers have become a source of interest for forensic scientists.The aim of this review is to consider the potential of alleles from the Human Leukocyte Antigen (HLA) complex as effective forensic markers. While Massively Parallel Sequencing of HLA is routinely used in clinical laboratories in fields such as transplantation, pharmacology or population studies, there have not been any studies testing its suitability for forensic casework samples.     

Evaluating the Probative Value of Child Custody Evaluations

It has previously been argued that a competent forensic work product is defined, in part, by the evaluator's use of conventional forensic methods and procedures applied to child custody evaluations (Gould, 1998) and that the more judges and other legal professionals understand about forensic methods and procedures, the better they are able to critically weigh the substance and merit of a child custody evaluation (Gould & Bell, 2000). These forensic methods and procedures have their foundation in the behavioral sciences and are characteristic of competent and comprehensive forensic evaluations conducted for other legal purposes. In this paper, we provide a more detailed model for critiquing the forensic competence of a child custody report. Such a model better assists courts and lawyers in understanding how to assess the substance and admissibility of custody reports.     

The accuracy of three methods of age estimation using radiographic measurements of developing teeth

The accuracy of age estimation using three quantitative methods of developing permanent teeth was investigated. These were M?rnstad et al. [Scand. J. Dent. Res. 102 (1994) 137], Liversidge and Molleson [J. For. Sci. 44 (1999) 917] and Carels et al. [J. Biol. Bucc. 19 (1991) 297]. The sample consisted of 145 white Caucasian children (75 girls, 70 boys) aged between 8 and 13 years. Tooth length and apex width of mandibular canine, premolars and first and second molars were measured from orthopantomographs using a digitiser. These data were substituted into equations from the three methods and estimated age was calculated and compared to chronological age. Age was under-estimated in boys and girls using all the three methods; the mean difference between chronological and estimated ages for method I was -0.83 (standard deviation +/-0.96) years for boys and -0.67 (+/-0.76) years for girls; method II -0.79 (+/-0.93) and -0.63 (+/-0.92); method III -1.03 (+/-1.48) and -1.35 (+/-1.11) for boys and girls, respectively. Further analysis of age cohorts, found the most accurate method to be method I for the age group 8.00-8.99 years where age could be predicted to 0.14+/-0.44 years (boys) and 0.10+/-0.32 years (girls). Accuracy was greater for younger children compared to older children and this decreased with age.     

Child and adolescent victims in forensic autopsy: a 5-year retrospective study

Children of more than 3 years of age and adolescents have been largely overlooked in the forensic literature, especially the 4-9 age group. Thus, the present study was undertaken to address this particular issue of child and adolescent victims in forensic autopsies. On a 5-year period (2000-2004) in Quebec province (Canada), all forensic autopsy cases of children and adolescents from 4 to 19 years of age were retrospectively studied. A total of 223 cases of child and adolescent deaths (148 males, 75 females; 6.6% of all forensic autopsies) were reviewed. Age, gender, manner of death and cause of death were analyzed for all victims organized into three groups of age: 4-9, 10-14, and 15-19. Moreover, homicide cases (n = 54) are further analyzed in terms of method(s) used (firearm, sharp force, asphyxia, blunt force, intoxication). This 5-year retrospective study may contribute to a better understanding of typical deaths in the 4-19 group of age and therefore, bring a working basis for the forensic pathologist or medical examiner/coroner.     

Highly sensitive screening method for nitroaromatic, nitramine and nitrate ester explosives by high performance liquid chromatography-atmospheric pressure ionization-mass spectrometry (HPLC-API-MS) in forensic applications

A highly sensitive screening method based on high performance liquid chromatography atmospheric pressure ionization mass spectrometry (HPLC-API-MS) has been developed for the analysis of 21 nitroaromatic, nitramine and nitrate ester explosives, which include the explosives most commonly encountered in forensic science. Two atmospheric pressure ionization (API) methods, atmospheric pressure chemical ionization (APCI) and electrospray ionization (ESI), and various experimental conditions have been applied to allow for the detection of all 21 explosive compounds. The limit of detection (LOD) in the full-scan mode has been found to be 0.012-1.2 ng on column for the screening of most explosives investigated. For nitrobenzene, an LOD of 10 ng was found with the APCI method in the negative mode. Although the detection of nitrobenzene, 2-, 3-, and 4-nitrotoluene is hindered by the difficult ionization of these compounds, we have found that by forming an adduct with glycine, LOD values in the range of 3-16 ng on column can be achieved. Compared with previous screening methods with thermospray ionization, the API method has distinct advantages, including simplicity and stability of the method applied, an extended screening range and a low detection limit for the explosives studied.