Artificial intelligence for imaging-based COVID-19 detection: Systematic review comparing added value of AI versus human readers

PURPOSE

A growing number of studies have examined whether Artificial Intelligence (AI) systems can support imaging-based diagnosis of COVID-19-caused pneumonia, including both gains in diagnostic performance and speed. However, what is currently missing is a combined appreciation of studies comparing human readers and AI.

METHODS

We followed PRISMA-DTA guidelines for our systematic review, searching EMBASE, PUBMED and Scopus databases. To gain insights into the potential value of AI methods, we focused on studies comparing the performance of human readers versus AI models or versus AI-supported human readings.

RESULTS

Our search identified 1270 studies, of which 12 fulfilled specific selection criteria. Concerning diagnostic performance, in testing datasets reported sensitivity was 42-100% (human readers, n = 9 studies), 60-95% (AI systems, n = 10) and 81-98% (AI-supported readers, n = 3), whilst reported specificity was 26-100% (human readers, n = 8), 61-96% (AI systems, n = 10) and 78-99% (AI-supported readings, n = 2). One study highlighted the potential of AI-supported readings for the assessment of lung lesion burden changes, whilst two studies indicated potential time savings for detection with AI.

CONCLUSIONS

Our review indicates that AI systems or AI-supported human readings show less performance variability (interquartile range) in general, and may support the differentiation of COVID-19 pneumonia from other forms of pneumonia when used in high-prevalence and symptomatic populations. However, inconsistencies related to study design, reporting of data, areas of risk of bias, as well as limitations of statistical analyses complicate clear conclusions. We therefore support efforts for developing critical elements of study design when assessing the value of AI for diagnostic imaging.

Assessment of the opportunities for increasing the availability of EU data on consumer product-related injuries

The availability of data on consumer products-related accidents and injuries is of interest to a wide range of stakeholders, such as consumer product safety and injury prevention policymakers, market surveillance authorities, consumer organisations, standardisation organisations, manufacturers and the public. While the amount of information available and potentially of use for product safety is considerable in some European Union (EU) countries, its usability at EU level is difficult due to high fragmentation of the data sources, the diversity of data collection methods and increasing data protection concerns. To satisfy the policy need for more timely information on consumer product-related incidents, apart from injury data that have been historically collected by the public health sector, a number of 'alternative' data sources were assessed as potential sources of interest. This study explores the opportunities for enhancing the availability of data of consumer product-related injuries, arising from selected existing and 'alternative' data sources, widely present in Europe, such as firefighters' and poison centres' records, mortality statistics, consumer complaints, insurance companies' registers, manufacturers' incident registers and online news sources. These data sources, coupled with the use of IT technologies, such as interlinking by remote data access, could fill in the existing information gap. Strengths and weaknesses of selected data sources, with a view to support a common data platform, are evaluated and presented. Conducting the study relied on the literature review, extensive use of the surveys, interviews, workshops with experts and online data-mining pilot study.

Chemical alternatives assessment of different flame retardants - A case study including multi-walled carbon nanotubes as synergist

Flame retardants (FRs) are a diverse group of chemicals used as additives in a wide range of products to inhibit, suppress, or delay ignition and to prevent the spread of fire. Halogenated FRs (HFRs) are widely used because of their low impact on other material properties and the low loading levels necessary to meet the required flame retardancy. Health and environmental hazards associated with some halogenated FRs have driven research for identifying safer alternatives. A variety of halogen-free FRs are available on the market, including organic (phosphorus and nitrogen based chemicals) and inorganic (metals) materials. Multi-walled carbon nanotubes (MWCNT) have been demonstrated to act as an effective/synergistic co-additive in some FR applications and could thereby contribute to reducing the loading of FRs in products and improving their performance. As part of the FP7 project DEROCA we carried out a chemical alternatives assessment (CAA). This is a methodology for identifying, comparing and selecting safer alternatives to chemicals of concern based on criteria for categorising human and environmental toxicity as well as environmental fate. In the project we assessed the hazard data of different halogen-free FRs to be applied in 5 industrial and consumer products and here we present the results for MWCNT, aluminium diethylphosphinate, aluminium trihydroxide, N-alkoxy hindered amines and red phosphorus compared to the HFR decabromodiphenylether. We consulted the REACH guidance, the criteria of the U.S.-EPA Design for Environment (DfE) and the GreenScreen® Assessment to assess and compare intrinsic properties affecting the hazard potential. A comparison/ranking of exposure reference values such as Derived No Effect Levels (DNELs) showed that FRs of concern are not identified by a low DNEL. A comparison based on hazard designations according to the U.S.-EPA DfE and GreenScreen® for human health endpoints, aquatic toxicity and environmental fate showed that the major differences between FRs of concern and their proposed alternatives are the potential for bioaccumulation and CMR (carcinogenic, mutagenic or reprotoxic) effects. As most alternatives are inorganic chemicals, persistence (alone) is not a suitable criterion. From our experiences in carrying out a CAA we conclude: i) REACH registration dossiers provide a comprehensive source of hazard information for an alternative assessment. It is important to consider that the presented data is subject to changes and its quality is variable. ii) Correct identification of the chemicals is crucial to retrieve the right data. This can be challenging for mixtures, reaction products or nanomaterials or when only trade names are available. iii) The quality of the data and the practice on how to fill data gaps can have a huge impact on the results and conclusions. iv) Current assessment criteria have mainly been developed for organic chemicals and create challenges when applied to inorganic solids, including nanomaterials. It is therefore crucial to analyse and report uncertainties for each decision making step.

Conservative management of a large occipital extradural haematoma

A 26-year-old male presented with a mild confusion and hemianopsia after traumatic brain injury. Cerebral CT-scan revealed a 62.5 cm(3) left occipital extradural haematoma (EDH). Although conventional neurosurgical management would have been to evacuate the haematoma, a conservative strategy was preferred, and the patient made a total recovery.

Risk management measures for chemicals in consumer products: documentation, assessment, and communication across the supply chain

This paper analyzes the way risk management measures (RMMs) for consumer products have been used to date in authority and industry risk assessments. A working concept for consumer product RMMs is developed, aimed at controlling, limiting or avoiding exposures, and helping to insure the safe use (or handling) of a substance as part of a consumer product. Particular focus is placed on new requirements introduced by REACH (registration, evaluation, and authorization of chemicals). A RMMs categorization approach is also developed, dividing consumer product RMMs into those that are product integrated and those that are communicated to consumers. For each of these categories, RMMs for normal use, accidental use or misuse need to be distinguished. The level of detail for documenting, assessing and communicating RMMs across supply chains can vary, depending on the type of the assessment (tiered approach). Information on RMMs was collected from published sources to demonstrate that a taxonomical approach using standard descriptors for RMMs libraries is needed for effective information exchange across supply chains.