Facial recognition for disaster victim identification.

Mass disaster events can result in high levels of casualties that need to be identified. Whilst disaster victim identification (DVI) relies on primary identifiers of DNA, fingerprints, and dental, these require ante-mortem data that may not exist or be easily obtainable. Facial recognition technology may be able to assist. Automated facial recognition has advanced considerably and access to ante-mortem facial images are readily available. Facial recognition could therefore be used to expedite the DVI process by narrowing down leads before primary identifiers are made available. This research explores the feasibility of using automated facial recognition technology to support DVI. We evaluated the performance of a commercial-off-the-self facial recognition algorithm on post-mortem images (representing images taken after a mass disaster) against ante-mortem images (representing a database that may exist within agencies who hold face databases for identity documents (such as passports or driver's licenses). We explored facial recognition performance for different operational scenarios, with different levels of face image quality, and by cause of death. Our research is the largest facial recognition evaluation of post-mortem and ante-mortem images to date. We demonstrated that facial recognition technology would be valuable for DVI and that the performance varies by image quality and cause of death. We provide recommendations for future research.

Bite marks: To opine or not to opine?

Much has been written about bite mark comparisons and bite mark analysis. It has largely been written for, and remained within the domain of, the forensic odontologist despite the limited number of such specialists and the even smaller subset who have expertise in this area. For those health professionals who work in the field of clinical forensic medicine, most will not have access to a forensic odontologist. Courts can be reticent about the costs involved in obtaining another expert opinion. For health professionals, who will likely encounter more bite marks in their career than forensic odontologists, the difficulty is knowing what can legitimately be opined about such injuries.

Impacts on paediatricians testifying in cases of child maltreatment: protocol for a systematic review.

INTRODUCTION: Paediatricians perform medical assessments for children in cases of suspected child maltreatment. Due to their role with statutory child protection agencies and police, paediatricians may be asked to testify in court about child protection and criminal justice matters. To the authors' knowledge, there has been no previous systematic review of the literature synthesising the evidence on the impacts on paediatricians testifying in cases of child maltreatment. METHODS AND ANALYSIS: A search strategy comprising indexed and key terms will be applied to six electronic reference databases from inception to May 2023: Medline, EMBASE, PsycINFO, CINAHL, Criminal Justice Abstracts and Cochrane Library. Two reviewers will independently screen titles and abstracts and full-text articles against predefined eligibility criteria to identify studies of interest. Conflicts will be independently adjudicated by a third reviewer. ETHICS AND DISSEMINATION: Since the systematic review methodology aims at synthesising information from available publications, this study does not require ethical approval. An article reporting the results of the systematic review will be submitted for publication in a scientific journal, presented at relevant conferences and used in subsequent stakeholder consultations.

Estimating dental age of New Zealand juveniles and subadults using Demirjian's method.

Reference data for dental age estimate is sparse in New Zealand (NZ), with only two contemporary studies. Te Moananui et al. (J For Sci. 53(2), 2008) presented modified Demirjian percentile curves to estimate dental age of Pasifika, Maori, and European males and females (n = 1383), while Timmins et al. (Forensic Sci Med Pathol. 8:101-8, 2012) found the Demirjian method (1973) was valid for a smaller sample (n = 200) of unknown ancestry. The study presented here sought to validate the Demirjian and the Te Moananui methods for a sample of the NZ population of unknown ancestry and a subgroup of known ancestry i.e., Pasifika, Maori and European, for males and females. The Demirjian method (1976) was applied to the current study's sample consisting of 3523 individuals aged 4 to < 20 years. The seven left mandibular teeth (third molar excluded) and tooth scores were summed for each individual, with the Te Moananui methods applied to this subgroup. The results revealed these methods to be less than ideal for estimating dental age of the NZ sample, for both males and females. The probit regression form of Transition Analysis (TA) was employed to calculate the mean age entering each tooth stage, for the seven teeth, to reduce age mimicry that is commonly associated with traditional regression analysis. TA results revealed Pasifika and Maori individuals to be more advanced than Caucasian individuals. The sex groups were also compared to the mean ages presented by Demirjian and Levesque with mixed results (J Dent Res. 59(7):1110-22, 1980), highlighting the need for more research in this area.

Effectiveness of low dose computed tomography to detect fractures in paediatric suspected physical abuse: a systematic review.

PURPOSE: The skeletal survey X-ray series is the current 'gold standard' when investigating suspected physical abuse (SPA) of children, in addition to a non-contrast computed tomography (CT) brain scan. This systematic literature review synthesised findings of published research to determine if low dose computed tomography (LDCT) could detect subtle fractures and therefore replace the skeletal survey X-ray series in the investigation of SPA in children aged under 3 years. METHODS: Five electronic databases and grey literature were systematically searched from their inception to 28 April 2022. Primary studies were included where the population comprised paediatric patients up to 16 years and LDCT was used to detect fractures associated with SPA. Studies involving imaging investigations of the head, standard dose CT examinations or accidental trauma were excluded. RESULTS: Three studies met the inclusion criteria, all of which were case series. These studies did not report many of the criteria required to compare the accuracy of LDCT to X-ray, i.e. they did not meet the criteria for a diagnostic accuracy test. Therefore, it is difficult to conclude from the case series if LDCT is accurate enough to replace X-rays. CONCLUSION: Due to the gap in current literature, a phantom study and subsequent post-mortem CT study are recommended as the primary investigative methods to assess the ability of low-dose CT to identify the subtle fractures associated with SPA and to calculate how low the achievable CT dose can be.

Pericardial Effusion Detection on Post-Mortem Computed Tomography Images Using Convolutional Neural Networks.

Pericardial effusion can be a sign of significant underlying diease and, in some cases, may lead to death. Post-mortem computed tomography (PMCT) is a well-established tool to assist death investigation processes in the forensic setting. In practice, the scarcity of well-trained radiologists is a challenge in processing raw whole-body PMCT images for pericardial effusion detection. In this work, we propose a Pericardial Effusion Automatic Detection (PEAD) framework to automatically process raw whole-body PMCT images to filter out the irrelevant images with heart organ absent and focus on pericardial effusion detection. In PEAD, the standard convolutional neural network architectures of VGG and ResNet are carefully modified to fit the specific characteristics of PMCT images. The experimental results prove the effectiveness of the proposed framework and modified models. The modified VGG and ResNet models achieved superior detection accuracy than the standard architecture with reduced processing speed.

Comparison of typical radiation doses and risks using an anthropomorphic 'bone fracture' phantom for commonly performed X-ray projections in a 5-year-old.

INTRODUCTION: Diagnostic reference levels (DRLs) are typical dose levels for medical imaging examinations for groups of standard-sized patients or standard phantoms for broadly defined types of equipment used as a tool to aid optimisation of protection for medical exposures. Currently, there are no paediatric DRLs for conventional radiography (i.e. general X-rays) published in Australia. The aim of this study was to establish typical radiation doses and risks that are representative of those delivered for commonly performed X-ray projections for a 5-year-old/20 kg child using a 5-year-old anthropomorphic 'bone fracture' phantom in three dedicated paediatric radiology departments in Victoria. METHODS: A total of 20 projection images were acquired for a standard 5-year-old/20 kg phantom using digital radiography X-ray equipment. The air kerma-area product (KAP) measured at each centre by a KAP metre, which was calibrated to a national primary standard, was considered to represent the median value for that centre for each X-ray projection. Organ doses and effective dose were estimated using PCXMC software, and risks of radiation-induced cancer and radiation-induced death were calculated based on the BEIR VII report. RESULTS: The typical doses for the individual X-ray projections ranged from 3 mGy•cm2 to 86 mGy•cm2 , whilst the effective doses ranged from 0.00004 to 0.07 mSv. The radiation risks were 'minimal' to 'negligible'. CONCLUSION: The estimation of typical radiation doses and associated risks for a 5-year-old/20 kg phantom study provides reference values for guidance and is a first step in assisting optimisation at other institutions until national DRLs, based on patient data from the clinical setting, are published.

Surformer: An interpretable pattern-perceptive survival transformer for cancer survival prediction from histopathology whole slide images.

BACKGROUND AND OBJECTIVE: High-resolution histopathology whole slide images (WSIs) contain abundant valuable information for cancer prognosis. However, most computational pathology methods for survival prediction have weak interpretability and cannot explain the decision-making processes reasonably. To address this issue, we propose a highly interpretable neural network termed pattern-perceptive survival transformer (Surformer) for cancer survival prediction from WSIs. METHODS: Notably, Surformer can quantify specific histological patterns through bag-level labels without any patch/cell-level auxiliary information. Specifically, the proposed ratio-reserved cross-attention module (RRCA) generates global and local features with the learnable prototypes (pglobal, plocals) as detectors and quantifies the patches correlative to each plocal in the form of ratio factors (rfs). Afterward, multi-head self&cross-attention modules proceed with the computation for feature enhancement against noise. Eventually, the designed disentangling loss function guides multiple local features to focus on distinct patterns, thereby assisting rfs from RRCA in achieving more explicit histological feature quantification. RESULTS: Extensive experiments on five TCGA datasets illustrate that Surformer outperforms existing state-of-the-art methods. In addition, we highlight its interpretation by visualizing rfs distribution across high-risk and low-risk cohorts and retrieving and analyzing critical histological patterns contributing to the survival prediction. CONCLUSIONS: Surformer is expected to be exploited as a useful tool for performing histopathology image data-driven analysis and gaining new insights for interpreting the associations between such images and patient survival states.

Targeting tumor heterogeneity: multiplex-detection-based multiple instance learning for whole slide image classification.

MOTIVATION: Multiple instance learning (MIL) is a powerful technique to classify whole slide images (WSIs) for diagnostic pathology. The key challenge of MIL on WSI classification is to discover the critical instances that trigger the bag label. However, tumor heterogeneity significantly hinders the algorithm's performance. RESULTS: Here, we propose a novel multiplex-detection-based multiple instance learning (MDMIL) which targets tumor heterogeneity by multiplex detection strategy and feature constraints among samples. Specifically, the internal query generated after the probability distribution analysis and the variational query optimized throughout the training process are utilized to detect potential instances in the form of internal and external assistance, respectively. The multiplex detection strategy significantly improves the instance-mining capacity of the deep neural network. Meanwhile, a memory-based contrastive loss is proposed to reach consistency on various phenotypes in the feature space. The novel network and loss function jointly achieve high robustness towards tumor heterogeneity. We conduct experiments on three computational pathology datasets, e.g. CAMELYON16, TCGA-NSCLC, and TCGA-RCC. Benchmarking experiments on the three datasets illustrate that our proposed MDMIL approach achieves superior performance over several existing state-of-the-art methods. AVAILABILITY AND IMPLEMENTATION: MDMIL is available for academic purposes at https://github.com/ZacharyWang-007/MDMIL.

Changing trends and characteristics of one punch deaths in Australia (2012-2018).

One-punch assaults also known as 'coward punches', are characterised by a single severe blow to the head causing the victim to lose consciousness, resulting in a secondary impact between the head and surrounding environment. Such impacts may result in brain injury leading to fatality or permanent neurological impairment. In a previous publication, there were 90 one punch deaths around Australia between 2000 and 2012, mostly involving young men drinking alcohol at a licensed venue at the weekend. This prompted a surge of public education and awareness campaigns around Australia, in addition to regulatory and legislative changes aimed at curbing social violence. This retrospective descriptive study aimed to examine one punch deaths since 2012 in Australia to determine if there has been a decrease in deaths, and whether the demographics and circumstances of these deaths have changed. A search of the National Coronial Information System was undertaken for all closed coronial cases between 1 January 2012 and 31 December 2018. Additional information was collected from medicolegal reports including toxicology, pathology and coronial findings. There were 80 one punch fatalities in Australia, almost exclusively involving males. The median age was 43.5 (range 18-71) years and there was a decreasing trend in the number of deaths annually. Most fatal assaults occurred in the state of New South Wales (28.8%) followed by Queensland (23.8%), and in metropolitan locations (64.6%) rather than regional areas (35.4%). Alcohol was the most commonly detected drug, found in 47 cases of the 71 cases where toxicology results were available (66%), with a median concentration of 0.14 and 0.19 g/100 mL in antemortem and postmortem samples, respectively (range 0.005-0.32 g/100 mL). Five deaths reported methylamphetamine, with THC detected in 21.1% of cases. Assaults more commonly occurred on a footpath or roadside (41.3%), followed by a home or dwelling (32.5%). 8.8% of assaults occurred inside hotels, bars or other licenced venues. Most transpired on a weekday, which differed from the pre-2012 period when these assaults occurred mainly on the weekend. While some trends are positive, there has been a shift in the victim demographic as well as the typical environment for fatal one punch assaults, highlighting the importance of public health surveillance in providing a current evidence base to inform policy and practice.

Myocarditis in the forensic setting - a review of the literature.

Diagnosis of myocarditis as the cause of death at post-mortem is currently determined by a forensic pathologist. There is no systematic method for diagnosis and thus the determination is subject to inter-observer variability and is non-reproducible. Postmortem studies often rely on the clinical method of diagnosis, which is inaccurate. Furthermore, there is no current standardized method of distinguishing between myocarditis as cause of death, and myocardial inflammation as an incidental finding post-mortem. Only a few studies have investigated a method of quantifying this difference using variables such as number of inflammatory cells and presence of myocyte necrosis, however, there are several limitations hindering the reproducibility of this research. This review investigates the current practices and limitations associated with the diagnosis of myocarditis as cause of death in the autopsy setting.

Myocarditis in the forensic setting.

Diagnosis of myocarditis as the cause of death in the forensic setting at post-mortem is currently determined by a forensic pathologist. There is no systematic method for diagnosis and thus the determination is subject to inter-observer variability and is often non-reproducible. The primary aim of this study was to investigate the differences in the amount of inflammation between cases where myocarditis was deemed the cause of death, compared to cases where myocardial inflammation was incidentally present at autopsy, but not determined to be the cause of death. Participants were sourced from the Victorian Institute of Forensic Medicine (VIFM) database, from full autopsies conducted on reportable death in Victoria, Australia between the years 2011 and 2021. Cases of fatal myocarditis were significantly more likely to experience infection-like symptoms prior to death, and to be in hospital at the time of death. Histopathological examination revealed fatal cases had a significantly higher inflammatory index compared to the incidental group. Lethal cases were also significantly more likely to have myocyte necrosis, and a diffuse pattern of inflammation. There are significant differences between cases where myocardial inflammation has been determined to be the cause of death and cases where inflammation in the myocardium was an incidental finding. These results could be used in the forensic autopsy to help pathologists determine if inflammation should be considered fatal or incidental.

Typical median effective radiation doses using an anthropomorphic bone fracture phantom for initial radiographic skeletal surveys in the investigation of suspected physical abuse.

BACKGROUND: A series of 31 radiographs is recommended by the Royal College of Radiologists (RCR) when investigating suspected physical abuse (SPA). OBJECTIVE: To determine the radiation dose delivered for skeletal surveys performed for SPA in Victorian radiology departments based on their local protocols. MATERIALS AND METHODS: A 5-year-old paediatric bone fracture phantom was radiographed at five radiology sites using both the RCR recommended protocol and, where applicable, the local departmental SPA protocol. The radiation doses were measured and recorded. These were scaled down to estimate the effective radiation doses for a 2-year-old child at each site and the associated radiation risks estimated. RESULTS: The median effective dose for all radiographic projections in the RCR skeletal survey radiographic series was 0.09 mSv. The estimated risk of radiation-induced cancer and radiation-induced death from cancer for 2-year-old children is classified as "very low," with girls having a higher risk than boys. CONCLUSION: The median effective radiation dose for the RCR skeletal survey for imaging in SPA was 0.09 mSv resulting in a "very low" additional risk of radiation-induced cancer. The authors will now aim to ascertain whether whole-body CT skeletal survey can replace the radiographic series for imaging in SPA while maintaining a comparable radiation dose.

Family violence homicide rates: a state-wide comparison of three data sources in Victoria, Australia.

BACKGROUND: Family violence homicide (FVH) is a major public health and social problem in Australia. FVH trend rates are key outcomes that determine the effectiveness of current management practices and policy directions. Data source-related methodological problems affect FVH research and policy and the reliable measurement of homicide trends. OBJECTIVE: This study aimed to determine data reliability and temporal trends of Victorian FVH rates and sex and relationship patterns. METHOD: FVH rates per 100,000 persons in Victoria were compared between the National Coronial Information System (NCIS), Coroners Court of Victoria (CCoV) Homicide Register, and the National Homicide Monitoring Program (NHMP). Trends for 2001-2017 were analysed using Joinpoint regression. Crude rates were determined by sex and relationship categories using annual frequencies and Australian Bureau of Statistics population estimates. RESULTS: NCIS closed FVH cases totalled 360, and an apparent downward trend in the FVH rate was identified. However, CCoV and NHMP rates trended upwards. While NCIS and CCoV were case-based, NHMP was incident-based, contributing to rate variations. The NCIS-derived trend was particularly impacted by unavailable case data, potential coding errors and entry backlog. Neither CCoV nor NHMP provided victim-age in their public domain data to enable age-adjusted rate comparison. CONCLUSION: Current datasets have limitations for FVH trend determination; most notably lag times for NCIS data. IMPLICATIONS: This study identified an indicative upward trend in FVH rates in Victoria, suggesting insufficiency of current management and policy settings for its prevention and control.

Exploring the use of machine learning for the assessment of skeletal fracture morphology and differentiation between impact mechanisms: A pilot study.

Analyzing and interpreting traumatic injuries is a fundamental aspect of routine forensic case work. As the human skeleton can be impacted through a combination of loading mechanisms and varying impact energies, the analysis and interpretation of skeletal trauma can be complex. Therefore, it is imperative that the reliability of techniques used for analysis are well-established. There is growing interest in machine learning (ML) in medicine (especially radiology) regarding the use of image classification (a subset of ML) to categorize and predict classes of medical images. Therefore, the feasibility of using image classification for skeletal trauma analysis should be explored for its benefits to forensic pathology and anthropology. The method explored in this paper examined the potential for machine learning, using three dimensional (3D) convolutional neural networks (CNNs), to assess whether morphological features of skeletal trauma to the femur can be used to differentiate between impact mechanisms within a forensic population. The objective of this study was to assess if morphological differences in femoral fractures seen in post-mortem-computed tomographic images (PMCT) could be categorized according to mechanism, specifically horizontal impacts resulting from pedestrian motor vehicle impacts (PMVIs) and vertical impact s resulting from high impact falls. Final model results indicated an accuracy between 69.95%-72.86% and 63.08%-66.24% validation. Although these results mean the method could not be practically used in its current form, as a proof of concept, there is potential for it to be developed as a tool to assist in classifying complex fracture states.

Technology-facilitated sexual assault in children and adolescents; is there a cause for concern? Fourteen years of experience at a metropolitan forensic paediatric medical service.

AIM: To quantify and characterise sexual assaults occurring after 12-17-year-old children connect with an alleged offender online (technology-facilitated sexual assault: TFSA) examined at a paediatric forensic medical service in Melbourne, Australia between 2014 and 2020. To compare these findings to a previous 7-year audit (2007-2013) to determine if there has been an increase in TFSA over the last 14 years. METHODS: A retrospective audit was undertaken of medical records of children aged 12-17 who underwent forensic medical examinations at the Victorian Forensic Paediatric Medical Service following an allegation of sexual assault, between 1 January 2014 and 31 December 2020. Cases identified as being technology facilitated were further analysed. Results from 2014 to 2020 7-year cohort were then compared to an existing audit over the preceding 7-year period (2007-2013). RESULTS: Of 515 sexual assault cases between 2014 and 2020, 70 (14%) victims reported connecting with the offender on a technological platform, compared to 4% of cases between 2007 and 2013. In 2019 and 2020, TFSA comprised almost one-fifth of caseload. Of the 70 TFSA cases, the majority occurred at the first face-to-face meeting following a variable period of online communication. Approximately one-third of TFSA's occurred at the offender's residence, and another third in a public place (park, public toilets). Technological platforms used evolved over the 14-year study period. CONCLUSIONS: The proportion of TFSA caseload seen at an Australian paediatric forensic medical service increased over the last 14 years, with common characteristics to these sexual assaults suggested.

Orofacial injuries in child family homicide: a population study.

Child family violence homicide (FVH) is a significant public health problem in Australia and globally. Population-wide studies of orofacial injuries in child FVH are uncommon despite their recognized importance. This whole population descriptive study of orofacial injuries in child FVH in Victoria, Australia aims to implement a novel methodological approach to provide an overview of child FVH and describe frequency and patterns of abusive orofacial injuries. Closed cases of child FVH aged 0-17 years, January 2000-December 2018, were identified from screening all Victorian assault deaths for eligible offender relationships. Significant associations of clinical/demographic characteristics were explored using two-step clustering and the Spearman correlation coefficient. Of 895 closed homicide cases, 358 were FV-related. Of the 53 child FVH, 40 were eligible for injury analysis with 36 of these cases (90%) having orofacial injuries. Among these 36 cases, 72% were aged 0-4 years, males predominated (64%) and the injury mechanism was blunt force for 56%. The discrete orofacial injury frequency was associated with the non-orofacial injury frequency (rho: 0.362, 2-tailed p < 0.03). A three-cluster statistical solution was identified, each represented by an injury mechanism. The largest cluster identified a pattern of blunt force trauma in 0-4 years with drug presence, high average non-orofacial injury numbers and parent-offenders. A novel methodological approach was implemented to comprehensively describe the frequency, nature, patterns and risk indicators of orofacial injuries in child FVH. It explored associations between a wide range of clinical and demographic characteristics, which might have otherwise been missed in summary description. These methods will potentially underpin future comparative studies of intentional-unintentional child injuries and fatal-nonfatal child abuse. The study narrows a significant research gap regarding patterns of inflicted injuries, and demographic and clinical indicators in child FVH potentially informing future systematic classification processes, risk assessment tools and pathways to FV intervention.

Analyzing the outcomes of skeletal trauma within a forensic population: Potential issues and implications in inferential modeling of blunt force trauma.

Analyzing complex skeletal trauma can present a challenge for forensic practitioners to reliably determine the causes and circumstances of traumatic injury. The forensic value of skeletal fracture pattern analysis can be diminished due to obscuration and similarity between injuries associated with various impact mechanisms (e.g., vertical vs horizontal blunt force) and can provide issues when questions arise surrounding circumstances of traumatic injuries. Using the Python coding language, code was written that segments traumatized regions of interest from post-mortem computed tomography (PMCT) scans, allowing the user to calculate percentage of fragmentation in the context of extreme trauma events. Using cases of known trauma which resulted in fragmentation of the femur, the objective was to assess if there were statistical differences in the fragmentation resulting from horizontal pedestrian motor vehicle impacts (PMVIs; n = 44) compared to vertical high impact falls (>3 m; n = 41). Results indicated that percentage differences between the PMVI group and high impact falls group were statistically significant. Although it was possible to develop a standardized method that records fragmentation, and results were significant in distinguishing between the two groups, the outcomes of the data follow an exponential distribution which has implications for how skeletal trauma is modeled.

The issues and complexities of establishing methodologies to differentiate between vertical and horizontal impact mechanisms in the analysis of skeletal trauma: An introductory femoral test.

Understanding skeletal trauma characteristics is fundamental for the examination and interpretation of blunt force trauma (BFT). BFT is the most complex type of trauma to interpret based on the analysis of skeletal fractures alone, with comminuted fractures presenting additional complications to assess and interpret. Considerable variation exists within each type of BFT injury dependent on direction, magnitude of force, plus a myriad of biological/environmental factors. Given the complex processes governing the nature of BFT skeletal injuries determining whether differences between impact mechanisms and skeletal trauma can be quantified requires investigation. AIM: this study aims to determine the feasibility of quantifying outcomes between two separate loading conditions by using a formula created from transformed variables recorded from specific trauma cases involving BFT to the femur. METHODOLOGY: Displacement, comminution, and femoral midshaft area data were recorded from full body postmortem computed tomography scans of 103 individuals (males, mean age 42.5, and females, mean age 48.9) where cause of death was the result of rapid horizontal deceleration impact events (pedestrian motor vehicular accidents, n = 59) and vertical (>3-metre falls, n = 44). These measurements were standardised and transformed into a continuous variable. Independent t-tests, binary logistic regression and K Nearest- Neighbours (KNN) were used to analyse the data. RESULTS: The standardised values showed mean group differences between falls (9.62) and pedestrian motor vehicular impacts (pedestrian MVAs) (9.53), however, these results were not statistically significant. The results indicate that similarities in variance between types of trauma outcomes and impact mechanisms demonstrate low equivalency (samples have limited differences), and the overall limitations in relying on using single elements to explain complex skeletal trauma outcomes.

Systematic Review of the Patterns of Orofacial Injuries in Physically Abused Children and Adolescents.

OBJECTIVE: To examine the evidence base for patterns of oro-facial injuries in physical abuse cases of children and adolescents aged 0-17 years. DATA SOURCES: Systematic searches of Ovid Medline, Ovid Embase, Cochrane Central and grey literature, dating from the oldest entry to August 2017. REVIEW METHODS: Search criteria included English language peer reviewed articles and theses on physical abuse cases affecting the age group of 0-17 years. Exclusion criteria were: Case reports and <10 case series; studies involving bite mark injuries elsewhere on the body, sexual, ritualistic or emotional abuse and neglect, exposure to domestic violence; reviews; book chapters and conference proceedings. The following data was extracted: quality and design of study, oro-facial manifestations, new/old injuries, fatalities, coexistent injuries, co-morbidities and radiologic investigations. Each study was subjected to two independent reviews and a third, if reviewers disagreed. RESULTS: The authors identified 51 articles, 26 of which satisfied the inclusion criteria. The oro-facial signs were superficial injuries of face, ears, neck, lips and oral mucosa, burns, torn fraenii and fractured teeth and jaws. A minority of studies stood out due to their well-developed design; expert opinion inclusion; new/old/occult injury investigations and facial bruising/ intraoral injuries as sentinel markers. Overall, the review demonstrated weak study quality and representativeness as well as lacunae in uniform reporting. CONCLUSIONS: The available literature does not endorse any pathognomonic signature patterns of child physical abuse to the exposed oro-facial region.