Machine learning algorithms have been drawing attention in lung disease research. However, due to their algorithmic learning complexity and the variability of their architecture, there is an ongoing need to analyze their performance. This study reviews the input parameters and the performance of machine learning applied to diagnosis of chronic obstructive pulmonary disease (COPD). One research focus of this study was on clearly identifying problems and issues related to the implementation of machine learning in clinical studies. Following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) protocol, 179, 1032, and 36,500 titles were identified from the PubMed, Scopus, and Google Scholar databases respectively. Studies that used machine learning to detect COPD and provided performance measures were included in our analysis. In the final analysis, 24 studies were included. The analysis of machine learning methods to detect COPD reveals the limited usage of the methods and the lack of standards that hinder the implementation of machine learning in clinical applications. The performance of machine learning for diagnosis of COPD was considered satisfactory for several studies; however, given the limitations indicated in our study, further studies are warranted to extend the potential use of machine learning to clinical settings.

Diagnostic medical imaging and the interpretation of the imaging results pose a great challenge for the medical profession as the final conclusions are highly susceptible to human error and subjectivity. The necessity for standardization of interpretation of medical images is very necessary to bypass these problems. The only way of achieving this is using a methodology which excludes the human eye and employs artificial intelligence. However, another challenge is selecting the most suitable AI algorithm fit for the challenging task of imaging results interpretation. This study was conducted following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines published in 2020. Research was done using PubMed, ScienceDirect and Google Scholar databases where the key inclusion criteria were language, journal credibility, open access to full-text publications and the most recent papers. In order to focus on only the most recent research, only the papers published in the last 5 years were evaluated. The search through PubMed, ScienceDirect and Google Scholar has yielded 81, 205, and 520 papers respectively. Out of this number of papers, 26 of them have met all of the inclusion criteria and were included in the research. The observed accuracies of the models and the overall rising interest in the topic denote that this field is rapidly growing and has a great potential to be applied in daily medical practice in the future.

Abstract Pneumonia is a leading cause of mortality in limited resource settings (LRS), which are common in low- and middle-income countries (LMICs). Accurate referrals can reduce the devastating impact of pneumonia, especially in LRS. Discriminating pneumonia from other respiratory conditions based only on symptoms is a major challenge. Machine learning has shown promise in overcoming the diagnostic difficulties of pneumonia (i.e., low specificity of symptoms, lack of accessible diagnostic tests and varied clinical presentation). Many scientific papers are now focusing on deep-learning methods applied to clinical images, which is unaffordable for initial patient referral in LMICs. The current study used a dataset of 4500 patients (1500 patients affected by bronchitis, 3000 by pneumonia) from a middle-income country, containing information on subject population characteristics, symptoms and laboratory test results. Manual feature selection was performed, focusing on clinical symptoms that are easily measurable in LRS and in community settings. Three common machine learning methods were tested and compared: logistic regression; decision tree and support vector machine. Models were developed through a holdout process of training-validation and testing. We focused on six clinically relevant, easily interpreted patient symptoms as best indicators for pneumonia. Our final model was a decision tree, achieving an AUC of 93%, with the advantage of being fully intelligible and easily interpreted. The performance achieved suggested that intelligible machine learning models can enhance symptom-based referral of pneumonia in LRS and in community settings.

The activities related to medical device market can be divided into pre-market and post-market surveillance. Pre-market processes have been defined by Medical Device Directives since 1992. These directives define all aspects of medical device design, production, testing, approval and certification. Appropriate standards have been adopted to support these activities. The CE mark issued by European Notified Bodies confirms that the medical device complies with safety and performance standards related to its class and that is therefore safe for intended usage. Post-market surveillance is not as well defined, so the new Medical Device Regulation addresses the identified gap and emphasizes the importance of standardizing and harmonizing the system for surveillance of medical devices already in use. The MDRs require stakeholders to monitor the quality, performance and safety of a device throughout its life cycle and to take corrective or preventive action when necessary. In this paper, we discuss the possibility of using artificial intelligence on Big Data structures resulting from the comprehensive methodology of post-market surveillance.

Artificial Intelligence (AI) has been drawing attention in the field of medical devices. However, due to system complexity, the variability of their architecture, as well as ethical and regulatory concerns there is an ongoing need to analyze its application and performance.This study presents a narrative commentary on the applications of artificial neural networks (ANN) and machine learning (ML) algorithms in medical devices, past, current and future perspectives of application. One research focus of this study was on identifying problems and issues related to the implementation of AI in medical devices. The commentary is based on scientific articles published in PubMed, Scopus ad ScienceDirect databases, official publications of international organizations: European Comission (EC), Food and Drug Administration (FDA), and World Health Organisation (WHO) published in 2009 - 2020 period. AI is revolutionizing healthcare, from medical applications to clinical engineering. However, before grasp-ing the full potential ethical, legal and social concerns need to be resolved and its application needs to be harmonized and regulated regarding equitable access, privacy, appropriate uses and users, liability and bias and inclusiveness.

The world is facing an unprecedented outbreak affecting all aspects of human lives which is caused by the COVID-19 pandemic. Due to the virus novelty, healthcare systems are challenged by a high rate of patients and the shortage of medical products. To address an increased need for essential medical products, national authorities, worldwide, made various legislative concessions. This has led to essential medical products being produced by automotive, textile and other companies from various industries and approved under the emergency use authorizations or legal concessions of national regulatory bodies. This paper presents a narrative commentary of the available documentation on emergency use authorizations and legal concessions for medical products during COVID-19 pandemic. The basis for narrative commentary includes scientific articles published in Web of Science, Scopus, PubMed and Embase databases, official publications of international organizations: Food and Drug Agency (FDA), World Health Organisation (WHO), World Bank and United Nations (UN), and national regulatory agency reports in native languages (English, German, Bosnian, and Croatian) published from November 1, 2019 to May 1, 2020. This paper focuses on three types of essential medical products: mechanical ventilators, personal protective equipment (PPE) and diagnostic tests. Evidence-informed commentary of available data and potential identified risks of emergency use authorizations and legal concessions is presented. It is recognized that now more than ever, raising global awareness and knowledge about the importance of respecting the essential requirements is needed to guarantee the appropriate quality, performance and safety of medical products, especially during outbreak situation, such as the COVID-19 pandemic. Emergency use authorizations for production, import and approval of medical products should be strictly specified and clearly targeted from case to case and should not be general or universal for all medical products, because all of them are associated with different risk level. Presented considerations and experiences should be taken as a guide for all possible future outbreak situations to prevent improvised reactions of national regulatory bodies.

Abstract Epilepsy is a neurological disorder characterised by unusual brain activity widely known as seizure affecting 4-7% of the world's population. The diagnosis of this disorder is currently based on analysis of the electroencephalography (EEG) signals in the time-frequency domain. The analysis is performed applying various algorithms that yield high performance, however the challenge of effective real-time epilepsy diagnosis persists. To address this, we have developed a Field Programmable Gate Array (FPGA) based solution for the classification of generalized and focal epileptic seizure types using a feed-forward multi-layer neural network architecture (MLP ANN). The neural network algorithm is trained, validated and tested on 822 captured signals from Temple University Hospital Seizure Detection Corpus (TUH EEG Corpus) database. Inputs into the system were five main features obtained from EEG signals by time-frequency analysis followed by Continuous Wavelet Transform (CWT) and subsequent statistical analysis. Out of the total number of samples, 583 (70 %) of them were utilised during the system development in MATLAB and TensorFlow and 239 (30 %) samples were further used for subsequent testing of the model performance on the FPGA. Subsequently, the adequate parameters of the ANN model were determined by using k-Fold Cross-Validation. Finally, the best performing ANN model in terms of average validation data accuracy achieved during cross-validation was implemented on the FPGA for real-time seizure classification. The digital ANN solution was coded in Very High-Speed Integrated Circuit Hardware Description Language (VHDL) and tested on the FPGA using 30 % reaming data. The results of this research demonstrate that epilepsy diagnosis with quite high accuracy (95.14 %) can be achieved with (5-12-3) MLP ANN implemented on FPGA. Also, the results show the steps towards appropriate implementation of ANN on the FPGA. These results can be utilised as the basis for the design of an application-specific integrated circuit (ASIC) allowing large serial production.

Microbiology laboratory is a type of medical laboratory and should be safe and efficient environment. Even it is not a mandatory for the accreditation in most of the countries, ISO/IEC 15189 remains the most common reference for quality of work in medical laboratories. It is mostly based on good laboratory practices and is oriented to support accurate clinical decisions. ISO/IEC 15189 has potential to become very effective instrument for development and improvement of medical laboratories. Results from laboratory should guide the majority of clinical decisions and help in providing adequate patient care. This article provides a simple approach to meet the minimum requirements set. To achieve intended goal and strictly follow the requirements proposed in the standard, the trained and well-motivated laboratory staff is necessary to implement the system. The objective of this article is for it to be used as a guideline for evaluation and implementation of the ISO 15189.

This paper presents the results of development of Artificial Neural Networks (ANNs) for prediction of medical device performance based on conformity assessment data. Conformity assessment data of medical devices was obtained from periodical inspections conducted by ISO 17020 accredited laboratory during 2015–2019 period. For the development of ANNs, 1738 samples of conformity assessment of infusion and perfusor pumps was used. Out of the overall number of samples, 1391 (80%) of them were used during system development and 346 (20%) samples were used for subsequent validation of system performance. During system development, the impact on overall system accuracy of different number of neurons in hidden layer and the activation functions was tested. Also, two neural network architectures were tested: feedforward and feedback. The results show that feedforward neural network architecture with 10 neurons in single hidden layer has the best performance. The overall accuracy of that neural network is 98.06% for performance prediction of perfusor pumps and 98.83% for performance prediction of infusion pumps. The recurrent neural network resulted in accuracy of 98.41% for both infusion pumps and perfusor pumps. The results show that conformity assessment data obtained through yearly inspections of medical devices can successfully be used for prediction of performance of single medical device. This is very important in increasing the safety and accuracy of diagnosis and treatment of patients.