Institut National de Police Scientifique

We propose a practical approach based on federated learning to solve out-of-domain issues with continuously running embedded speech-based models such as wake word detectors. We conduct an extensive empirical study of the federated averaging algorithm for the "Hey Snips" wake word based on a crowdsourced dataset that mimics a federation of wake word users. We empirically demonstrate that using an adaptive averaging strategy inspired from Adam in place of standard weighted model averaging highly reduces the number of communication rounds required to reach our target performance. The associated upstream communication costs per user are estimated at 8 MB, which is a reasonable in the context of smart home voice assistants. Additionally, the dataset used for these experiments is being open sourced with the aim of fostering further transparent research in the application of federated learning to speech data.

Errors in sample handling or test interpretation may cause false positives in forensic DNA testing. This article uses a Bayesian model to show how the potential for a false positive affects the evidentiary value of DNA evidence and the sufficiency of DNA evidence to meet traditional legal standards for conviction. The Bayesian analysis is contrasted with the "false positive fallacy," an intuitively appealing but erroneous alternative interpretation. The findings show the importance of having accurate information about both the random match probability and the false positive probability when evaluating DNA evidence. It is argued that ignoring or underestimating the potential for a false positive can lead to serious errors of interpretation, particularly when the suspect is identified through a "DNA dragnet" or database search, and that ignorance of the true rate of error creates an important element of uncertainty about the value of DNA evidence.

AIMS: To assess the trends in the number, mortality and the nature of forensic cases involving toxicological detection of buprenorphine or methadone among toxicological investigations performed in Paris from June 1997 to June 2002. DESIGN: Retrospective, 5 year study with review of premortem data, autopsy, police reports, hospital data, and post-mortem toxicological analyses. SETTING AND PARTICIPANTS: 34 forensic cases of buprenorphine and 35 forensic cases of methadone detection among 1600 toxicological investigations performed at the Laboratory of Toxicology in the Medical Examiner's Office in Paris. MEASUREMENTS AND RESULTS: Therapeutic, toxic or lethal drug concentrations were defined based upon the results of blood analyses and the published literature. Drug concentrations were cross-referenced with other available ante- and post-mortem data. Subsequently, we classified a 'clear responsibility', 'possible responsibility' or 'not causative' role for buprenorphine or methadone in the death process, or 'no explanation of death'. Buprenorphine and methadone can be regarded as being directly implicated in, respectively, four of 34 death cases (12%) and three of 35 death cases (9%), and their participation in the lethal process is strongly plausible in eight (buprenorphine) and 11 (methadone) additional deaths. CONCLUSIONS: Analysis of causes of death reveals the difficulties in determining the role of substitution drugs in the death process, as many other factors may be involved, including circumstances surrounding death, past history, differential selection of subjects into either substitution modality and concomitant intake of other drugs (especially benzodiazepines and neuroleptics). The potential for synergistic or additive actions by other isolated molecules-particularly opioids, benzodiazepines, other psychotropes and alcohol-must be also considered.

Determining the number of contributors to a forensic DNA mixture using maximum allele count is a common practice in many forensic laboratories. In this paper, we compare this method to a maximum likelihood estimator, previously proposed by Egeland et al., that we extend to the cases of multiallelic loci and population subdivision. We compared both methods' efficiency for identifying mixtures of two to five individuals in the case of uncertainty about the population allele frequencies and partial profiles. The proportion of correctly resolved mixtures was >90% for both estimators for two- and three-person mixtures, while likelihood maximization yielded success rates 2- to 15-fold higher for four- and five-person mixtures. Comparable results were obtained in the cases of uncertain allele frequencies and partial profiles. Our results support the use of the maximum likelihood estimator to report the number of contributors when dealing with complex DNA mixtures.

BACKGROUND: The objective of the study was to assess the impact of the relocation of an adult hematological intensive care unit on invasive aspergillosis (IA) incidence. METHODS: A quasi-experimental study, including a control group and an intervention group that both underwent pretest and posttest evaluations, was conducted in the 3 adult hematological intensive care units (each composed of 14 single rooms) in a university hospital from 14 April 2005 through 1 February 2006. One of these units was relocated from the main building to an adjoining modular construction. In this unit, 4 rooms were equipped with laminar airflow before relocation; all rooms were equipped with positive pressure isolation after relocation. The 2 other units (control group), each containing 8 rooms with laminar airflow, did not undergo environmental modification. The diagnostic criteria for IA were based on the criteria of the European Organization for Research and Treatment of Cancer. RESULTS: In total, 356 hospitalized patients were included. Of the 21 cases of IA, 18 were nosocomial, and 3 were of undetermined origin. In the relocated unit, the incidence of IA decreased from 13.2% (9 patients) before relocation to 1.6% (1 patient) after relocation (P=.018). Eight of the 9 patients with IA before relocation stayed in rooms without specific air treatment. The rate of IA did not change in the control group. Patient characteristics were similar in each unit before and after relocation. CONCLUSION: We detected a straightforward association between environmental modification and decreased IA incidence, which emphasizes the use of an environmental strategy, including high-efficiency air filtration, in the prevention of IA.

The use of earmarks as evidence in criminal trials appears to be expanding, but there is something of a dearth of peer-reviewed scientific publications to support the pursuit. This paper is a critical review of the current literature in which we emphasize the weaknesses of the present state of knowledge. Some research directions are proposed to gather statistical knowledge of the within-source and between-source variability of earmarks and earprints. Its ultimate goal is to be able to assess likelihood ratios in relation to this type of evidence.

The spread of disease through human populations is complex. The characteristics of disease propagation evolve with time, as a result of a multitude of environmental and anthropic factors, this non-stationarity is a key factor in this huge complexity. In the absence of appropriate external data sources, to correctly describe the disease propagation, we explore a flexible approach, based on stochastic models for the disease dynamics, and on diffusion processes for the parameter dynamics. Using such a diffusion process has the advantage of not requiring a specific mathematical function for the parameter dynamics. Coupled with particle MCMC, this approach allows us to reconstruct the time evolution of some key parameters (average transmission rate for instance). Thus, by capturing the time-varying nature of the different mechanisms involved in disease propagation, the epidemic can be described. Firstly we demonstrate the efficiency of this methodology on a toy model, where the parameters and the observation process are known. Applied then to real datasets, our methodology is able, based solely on simple stochastic models, to reconstruct complex epidemics, such as flu or dengue, over long time periods. Hence we demonstrate that time-varying parameters can improve the accuracy of model performances, and we suggest that our methodology can be used as a first step towards a better understanding of a complex epidemic, in situation where data is limited and/or uncertain.

A collaborative study on Raman spectroscopy was carried out by members of the ENFSI (European Network of Forensic Science Institutes) European Fibres Group (EFG) on three dyed fibers: two red acrylics and one red wool. Raman instruments from six different manufacturers were tested as well as nine different laser wavelengths ranging from blue (lambda = 458 nm) to near infrared-NIR (lambda = 1064 nm). This represents the largest comparison study of Raman analytical parameters carried out on identical fiber samples. For the chosen fiber and dye samples, red lasers (lambda = 633 and 685 nm) gave the poorest spectral quality whereas blue (458 nm), green (514 nm) and near infrared lasers (785, 830 and 1064 nm) provided average results. Blue (488 nm) and green lasers (532 nm) globally gave the best quality spectra. Fluorescence problems were often encountered with some of the excitation wavelengths and therefore a flexible Raman instrument equipped with different lasers can be recommended to measure forensic fiber samples. The instrument should also be equipped with a Raman microscope in order to be able to focus on a single fiber. This study shows that Raman spectroscopy usually enables the identification of the main dye present in a colored fiber; however, minor dye components are much more difficult to detect. SERRS (Surface Enhanced Resonance Raman Scattering) techniques give an improvement of the dye's spectral intensity but no spectral improvement was observed for the two red acrylic and red wool fibers tested.

Human scent identification is based on a matching-to-sample task in which trained dogs are required to compare a scent sample collected from an object found at a crime scene to that of a suspect. Based on dogs' greater olfactory ability to detect and process odours, this method has been used in forensic investigations to identify the odour of a suspect at a crime scene. The excellent reliability and reproducibility of the method largely depend on rigor in dog training. The present study describes the various steps of training that lead to high sensitivity scores, with dogs matching samples with 90% efficiency when the complexity of the scents presented during the task in the sample is similar to that presented in the in lineups, and specificity reaching a ceiling, with no false alarms in human scent matching-to-sample tasks. This high level of accuracy ensures reliable results in judicial human scent identification tests. Also, our data should convince law enforcement authorities to use these results as official forensic evidence when dogs are trained appropriately.

Studies in mice and guinea pigs have shown that albumin is a new biomarker of organophosphorus toxicant (OP) and nerve agent exposure. Our goal was to determine whether OP-labeled albumin could be detected in the blood of humans exposed to OP. Blood from four OP-exposed patients was prepared for mass spectrometry analysis by digesting 0.010 ml of serum with pepsin and purifying the labeled albumin peptide by offline high performance liquid chromatography. Dimethoxyphosphate-labeled tyrosine 411 was identified in albumin peptides VRY(411)TKKVPQVSTPTL and LVRY(411)TKKVPQVSTPTL from two patients who had attempted suicide with dichlorvos. The butyrylcholinesterase activity in these serum samples was inhibited 80%. A third patient whose serum BChE activity was inhibited 8% by accidental inhalation of dichlorvos had undetectable levels of adduct on albumin. A fourth patient whose BChE activity was inhibited 60% by exposure to chlorpyrifos had no detectable adduct on albumin. This is the first report to demonstrate the presence of OP-labeled albumin in human patients. It is concluded that tyrosine 411 of human albumin is covalently modified in the serum of humans poisoned by dichlorvos and that the modification is detectable by mass spectrometry. The special reactivity of tyrosine 411 with OP suggests that other proteins may also be modified on tyrosine. Identification of other OP-modified proteins may lead to an understanding of neurotoxic symptoms that appear long after the initial OP exposure.

Abstract The principal aim of this study is to determine why police officers are generally found to be more prejudiced towards disadvantaged groups than are the standard population. Two independent processes were expected to account for this effect: Selection and group socialisation. Using a cross‐sectional design (N = 170), firstly, we compared, newly recruited police officers with a control population (selection effect), and secondly, police officers with 1 year of training with the newly recruited ones (group socialisation effect). Results reveal a significant effect of both selection and group socialisation, the two being underlined by distinct processes; right‐wing authoritarianism (RWA) in the case of the former and internalisation of a prejudice norm in the case of the latter. Finally, the results show that group identification moderates the change in internalisation of the prejudice norm. Copyright © 2009 John Wiley & Sons, Ltd.

Capillary electrophoresis (CE) appeared as an interesting alternative to chromatographic methods for carbohydrate analysis, but it can be difficult to implement, because of the lack of easily ionizable functions and chromophore groups. Recently, a promising method was proposed by Rovio et al. for the CE separation under extremely high alkaline conditions of neutral carbohydrates under their alcoholate form and their direct UV detection [Rovio et al. Electrophoresis 2007, 28, 3129-3135; and Rovio et al. J. Chromatogr. A 2008, 1185, 139-144], which is claimed to be due to the absorption of enediolate at 270 nm. Even so, most of the detected compounds in Rovio's paper (for example, sucrose) cannot give such enediolate, lacking a carbonyl group. In this work, a deeper insight was paid to the understanding of detection mechanism. In effect, unusual detection phenomena were observed in comparing reducing and nonreducing carbohydrate behaviors, which pointed to the existence of photochemical reactions in the detection window. A more systematic study of the influence of many parameters (carbohydrate nature, electrolyte pH, residence time in the detection window, and capillary diameter) was undertaken. In addition to this, most of this work was performed under cathodic (reversed) electro-osmotic flow conditions (using Polybrene-modified capillaries), to obtain much faster separations than under Rovio's conditions. This study also opens up new avenues for the detection in mid-UV range of non-UV-absorbing compounds bearing reducing moieties, such as amino acids.

Summary Glutamate (Glu) is the cornerstone of nitrogen assimilation and photorespiration in illuminated leaves. Despite this crucial role, our knowledge of the flux to Glu de novo synthesis is rather limited. Here, we used isotopic labelling with 13 CO 2 and 13 C‐ NMR analyses to examine the labelling pattern and the appearance of multi‐labelled species of Glu molecules to trace the origin of C‐atoms found in Glu. We also compared this with 13 C‐labelling patterns in Ala and Asp, which reflect citrate (and thus Glu) precursors, that is, pyruvate and oxaloacetate. Glu appeared to be less 13 C‐labelled than Asp and Ala, showing that the Glu pool was mostly formed by ‘old’ carbon atoms. There were modest differences in intramolecular 13 C– 13 C couplings between Glu C‐2 and Asp C‐3, showing that oxaloacetate metabolism to Glu biosynthesis did not involve C‐atom redistribution by the Krebs cycle. The apparent carbon allocation increased with carbon net photosynthesis. However, when expressed relative to CO 2 fixation, it was clearly higher at low CO 2 while it did not change in 2% O 2 , as compared to standard conditions. We conclude that Glu production from current photosynthetic carbon represents a small flux that is controlled by the gaseous environment, typically upregulated at low CO 2 .

Abstract Hairs from cats and dogs may be extremely important when used as evidence in the investigation of certain crimes and offenses. Certain parameters such as transfer and persistence are likely to be very similar to those encountered with fibers. These two parameters have been examined for dog and cat hair. The results confirm that hairs cling easily to various surfaces and that they are transferred in large numbers. It is almost impossible to enter a house where a domestic animal lives without being contaminated by its hair.

Invasive aspergillosis (IA) is an important cause of morbidity and mortality in neutropenic patients with hematological malignancies. To investigate the immediate and mid-term benefits of posaconazole prophylaxis in AML patients undergoing first induction chemotherapy and to study the infection risk factors, we prospectively studied the IA incidence in these patients at our hospital between years 2007 and 2008; then we compared them to a matched control group without prophylaxis. There were 55 and 66 patients in each group respectively. At day 32 post-induction, two probable cases (3.6%) were scored in the prophylaxis group compared to 8 cases (12.1%) in the control group (4 possible and 4 probable). At day 100, it reached 7.27% and 15.5% respectively. Kaplan-Meier analysis at day 100 showed lower mortality rate in the prophylaxis group compared to the control group [3.64% (n = 2, none due to IA) and 10.61% (n = 7, four due to IA) respectively, P = 0.002]. Multivariate analysis showed age and lack of response to induction as independent infection risk factors. Posaconazole prophylaxis resulted in lower incidence of IA and significantly improved survival. Patient's age and response to induction treatment are two independent infection risk factors, and need more attention during future clinical trials linked to antifungal prophylaxis.

Journal Article THE EPIDEMIOLOGY OF BURKITT'S LYMPHOMA: EVIDENCE FOR A CAUSAL ASSOCIATION WITH EPSTEIN-BARR VIRUS Get access GUY DE-THE GUY DE-THE 2 Director of Research, Centre National de la Recherche Scientifique, University of Lyon, Faculty of Medicine Alexis CarrelLyon, France , and Institut de Recherches Scientifiques sur le CancerVillejuif, France Search for other works by this author on: Oxford Academic PubMed Google Scholar Epidemiologic Reviews, Volume 1, Issue 1, 1979, Pages 32–54, https://doi.org/10.1093/oxfordjournals.epirev.a036213 Published: 01 March 1979

Abstract The applicability of isotopic analysis of 13C for forensic purposes has been discussed in the case of heroin. The obtained results showed first that morphine acetylation induces an isotopic fractionation, so that the 13C enrichment of diacetylmorphine depends on both geographical origin of the sample and source of acetic anhydride used by the drug trafficker. That measurement can then be of great interest in the determination of common-batch samples (tactical intelligence purpose). Second, diacetylmorphine deacetylation has also been studied and it appeared that this deacetylation allows to relieve the enrichment from the acetylation-induced part. Therefore, measuring morphine 13C enrichment, from deacetylated heroin samples, can be useful for determining the geographical origin of the samples (strategic intelligence purpose). Moreover, measuring both diacetylmorphine and morphine 13C enrichments may help identify the source of acetic anhydride used by the drug trafficker, so that the fraudulent laboratory where the morphine samples have been acetylated.

AIMS: (1). To assess the trends in the number, mortality and the nature of severe opiate/opioid poisonings from 1995 to 1999 in north-east Paris and adjacent suburbs and (2). to examine the effects of the introduction of high-dose buprenorphine on these parameters. DESIGN: Retrospective, 5-year study with review of pre-hospital, hospital and post-mortem data. SETTING AND PARTICIPANTS: Eighty patients from the toxicological intensive care unit (TICU) in north-east Paris, 421 patients from the pre-hospital emergency medical service in a north-east suburb of Paris (SAMU 93) and 40 deaths from the coroner's office in Paris. MEASUREMENTS AND RESULTS: We found that the number of pre-hospital opiate/opioid poisonings and deaths decreased over 5 years. During the same time frame, opiate/opioid poisoning admissions to our TICU remained steady, but the number of deaths declined. From 1995 to 1999, the detection of buprenorphine among opiate/opioid-poisoned TICU patients increased from two to eight occurrences per year while detection of opiates diminished from 17 to 10 occurrences per year. Increased buprenorphine detection correlated directly with increasing sales over this time period. In spite of the increased use of buprenorphine, the mortality associated with opiate/opioid poisonings has diminished in the pre-hospital environment from 9% in 1995 to 0% in 1999, and in the TICU from 12% in 1995 to 0% in 1997 and thereafter. We found a high frequency of multiple opiate/opioid use in severe poisonings, as well as the frequent association of other psychoactive drugs including ethanol. CONCLUSIONS: The number and the mortality of opiate/opioid poisonings appear to be stable or decreasing in our region. The association of multiple opiates/ opioids appears nearly as common as the association with other psychoactive drugs. The introduction of high-dose buprenorphine coincides with a decrease in opiate/opioid poisoning mortality. Further study will be necessary to clarify this observation.

Carbofuran is a pesticide whose acute toxicity is due to inhibition of acetylcholinesterase. Butyrylcholinesterase (BChE) in plasma is inhibited by carbofuran and serves as a biomarker of poisoning by carbofuran. The goal was to develop a method to positively identify poisoning by carbofuran. Sera from an attempted murder and an attempted suicide were analyzed for the presence of carbofuran adducts on BChE. The BChE from 1 ml of serum was rapidly purified on a 0.2 ml procainamide-Sepharose column. Speed was essential because the carbofuran-BChE adduct decarbamylates with a half-life of about 2 h. The partially purified BChE was boiled to denature the protein, thus stopping decarbamylation and making the protein vulnerable to digestion with trypsin. The labeled peptide was partially purified by HPLC before analysis by LC/MS/MS in the multiple reaction monitoring mode on the QTRAP 2000 mass spectrometer. Carbofuran was found to be covalently bound to Ser 198 of human BChE in serum samples from two poisoning cases. Multiple reaction monitoring triggered MS/MS spectra positively identified the carbofuran-BChE adduct. In conclusion a mass spectrometry method to identify carbofuran poisoning in humans has been developed. The method uses 1 ml of serum and detects low-level exposure associated with as little as 20% inhibition of plasma butyrylcholinesterase.

The aim of this project was to investigate the evidential value of blue gel pen inks in Europe. For this purpose, 33 blue gel pen inks, of different brands and models, representative of those available on the European market at the time of the study, were analyzed using three techniques: filtered light examination (FLE), Raman Spectroscopy (RS), and scanning electron microscopy (SEM). First, after visual examination (naked eye and stereo microscope), it was possible to classify the 33 inks into three groups described as: milky, metallic, and normal. This paper describes in detail the results obtained for the normal gel pen group. The ability of the techniques to discriminate gel inks between and within brands varied. The results indicated that RS and SEM were more discriminating than FLE. The greatest degree of differentiation was achieved when using a combination of RS and SEM techniques (discriminating power = 0.91). This study also highlights some problems concerning the identification of the brand of a gel pen from a written text.