A two-staged classifier to reduce false positives: On device detection of atrial fibrillation using phase-based distribution of poincaré plots and deep learning.

BACKGROUND: Mobile Cardiac Outpatient Telemetry (MCOT) can be used to screen high risk patients for atrial fibrillation (AF). These devices rely primarily on algorithmic detection of AF events, which are then stored and transmitted to a clinician for review. It is critical the positive predictive value (PPV) of MCOT detected AF is high, and this often leads to reduced sensitivity, as device manufacturers try to limit false positives. OBJECTIVE: The purpose of this study was to design a two stage classifier using artificial intelligence (AI) to improve the PPV of MCOT detected atrial fibrillation episodes whilst maintaining high levels of detection sensitivity. METHODS: A low complexity, RR-interval based, AF classifier was paired with a deep convolutional neural network (DCNN) to create a two-stage classifier. The DCNN was limited in size to allow it to be embedded on MCOT devices. The DCNN was trained on 491,727 ECGs from a proprietary database and contained 128,612 parameters requiring only 158 KB of storage. The performance of the two-stage classifier was then assessed using publicly available datasets. RESULTS: The sensitivity of AF detected by the low complexity classifier was high across all datasets (>93%) however the PPV was poor (<76%). Subsequent analysis by the DCNN increased episode PPV across all datasets substantially (>11%), with only a minor loss in sensitivity (<5%). This increase in PPV was due to a decrease in the number of false positive detections. Further analysis showed that DCNN processing was only required on around half of analysis windows, offering a significant computational saving against using the DCNN as a one-stage classifier. CONCLUSION: DCNNs can be combined with existing MCOT classifiers to increase the PPV of detected AF episodes. This reduces the review burden for physicians and can be achieved with only a modest decrease in sensitivity.

Device agnostic AI-based analysis of ambulatory ECG recordings.

Deep Convolutional Neural Networks (DCNNs) have been shown to provide improved performance over traditional heuristic algorithms for the detection of arrhythmias from ambulatory ECG recordings. However, these DCNNs have primarily been trained and tested on device-specific databases with standardized electrode positions and uniform sampling frequencies. This work explores the possibility of training a DCNN for Atrial Fibrillation (AF) detection on a database of single­lead ECG rhythm strips extracted from resting 12­lead ECGs. We then test the performance of the DCNN on recordings from ambulatory ECG devices with different recording leads and sampling frequencies. We developed an extensive proprietary resting 12­lead ECG dataset of 549,211 patients. This dataset was randomly split into a training set of 494,289 patients and a testing set of the remaining 54,922 patients. We trained a 34-layer convolutional DCNN to detect AF and other arrhythmias on this dataset. The DCNN was then validated on two Physionet databases commonly used to benchmark automated ECG algorithms (1) MIT-BIH Arrhythmia Database and (2) MIT-BIH Atrial Fibrillation Database. Validation was performed following the EC57 guidelines, with performance assessed by gross episode and duration sensitivity and positive predictive value (PPV). Finally, validation was also performed on a selection of rhythm strips from an ambulatory ECG patch that a committee of board-certified cardiologists annotated. On MIT-BIH, The DCNN achieved a sensitivity of 100% and 84% PPV in detecting episodes of AF. and 100% sensitivity and 94% PPV in quantifying AF episode duration. On AFDB, The DCNN achieved a sensitivity of 94% and PPV of 98% in detecting episodes of AF, and 98% sensitivity and 100% PPV in quantifying AF episode duration. On the patch database, the DCNN demonstrated performance that was closely comparable to that of a cardiologist. The results indicate that DCNN models can learn features that generalize between resting 12­lead and ambulatory ECG recordings, allowing DCNNs to be device agnostic for detecting arrhythmias from single­lead ECG recordings and enabling a range of clinical applications.

Machine learning and the electrocardiogram over two decades: Time series and meta-analysis of the algorithms, evaluation metrics and applications.

BACKGROUND: The application of artificial intelligence to interpret the electrocardiogram (ECG) has predominantly included the use of knowledge engineered rule-based algorithms which have become widely used today in clinical practice. However, over recent decades, there has been a steady increase in the number of research studies that are using machine learning (ML) to read or interrogate ECG data. OBJECTIVE: The aim of this study is to review the use of ML with ECG data using a time series approach. METHODS: Papers that address the subject of ML and the ECG were identified by systematically searching databases that archive papers from January 1995 to October 2019. Time series analysis was used to study the changing popularity of the different types of ML algorithms that have been used with ECG data over the past two decades. Finally, a meta-analysis of how various ML techniques performed for various diagnostic classifications was also undertaken. RESULTS: A total of 757 papers was identified. Based on results, the use of ML with ECG data started to increase sharply (p < 0.001) from 2012. Healthcare applications, especially in heart abnormality classification, were the most common application of ML when using ECG data (p < 0.001). However, many new emerging applications include using ML and the ECG for biometrics and driver drowsiness. The support vector machine was the technique of choice for a decade. However, since 2018, deep learning has been trending upwards and is likely to be the leading technique in the coming few years. Despite the accuracy paradox, accuracy was the most frequently used metric in the studies reviewed, followed by sensitivity, specificity, F1 score and then AUC. CONCLUSION: Applying ML using ECG data has shown promise. Data scientists and physicians should collaborate to ensure that clinical knowledge is being applied appropriately and is informing the design of ML algorithms. Data scientists also need to consider knowledge guided feature engineering and the explicability of the ML algorithm as well as being transparent in the algorithm's performance to appropriately calibrate human-AI trust. Future work is required to enhance ML performance in ECG classification.

Home-based digital health technologies for older adults to self-manage multiple chronic conditions: A data-informed analysis of user engagement from a longitudinal trial.

Background: Ageing populations are resulting in higher prevalence of people with multiple chronic conditions (multimorbidity). Digital health platforms have great potential to support self-management of multimorbidity, increasing a person's awareness of their health and well-being, supporting a better understanding of diseases and encouraging behaviour change. However, little research has explored the long-term engagement of older adults with such digital interventions. Methods: The aim of this study is to analyse how 60 older adults with multimorbidity engaged with digital symptom and well-being monitoring through a digital health platform over a period of approximately 12 months. Data analysis focused on user retention, frequency of monitoring, intervals in monitoring and patterns of daily engagement. Results: Our findings show that the overall engagement with the digital health platform was high, with more than 80% of participants using the technology devices for over 200 days. The submission frequency for symptom parameters (e.g. blood glucose (BG), blood pressure (BP), etc.) was between three and four times per week which was higher than that of self-report (2.24) and weight (2.84). Submissions of exercise (6.12) and sleep (5.67) were more frequent. The majority of interactions happened in the morning time. The most common time of submission for symptom parameters was 10 am, whereas 8 am was the most common time for weight measurements. Conclusions: The findings indicate the patterns of engagement of older adults with complex chronic diseases with digital home-based self-management systems.

Remote synchronous usability testing of public access defibrillators during social distancing in a pandemic.

Public access automated external defibrillators (AEDs) represent emergency medical devices that may be used by untrained lay-persons in a life-critical event. As such their usability must be confirmed through simulation testing. In 2020 the novel coronavirus caused a global pandemic. In order to reduce the spread of the virus, many restrictions such as social distancing and travel bans were enforced. Usability testing of AEDs is typically conducted in-person, but due to these restrictions, other usability solutions must be investigated. Two studies were conducted, each with 18 participants: (1) an in-person usability study of an AED conducted in an office space, and (2) a synchronous remote usability study of the same AED conducted using video conferencing software. Key metrics associated with AED use, such as time to turn on, time to place pads and time to deliver a shock, were assessed in both studies. There was no difference in time taken to turn the AED on in the in-person study compared to the remote study, but the time to place electrode pads and to deliver a shock were significantly lower in the in-person study than in the remote study. Overall, the results of this study indicate that remote user testing of public access defibrillators may be appropriate in formative usability studies for determining understanding of the user interface.

System Usability Scale Benchmarking for Digital Health Apps: Meta-analysis.

BACKGROUND: The System Usability Scale (SUS) is a widely used scale that has been used to quantify the usability of many software and hardware products. However, the SUS was not specifically designed to evaluate mobile apps, or in particular digital health apps (DHAs). OBJECTIVE: The aim of this study was to examine whether the widely used SUS distribution for benchmarking (mean 68, SD 12.5) can be used to reliably assess the usability of DHAs. METHODS: A search of the literature was performed using the ACM Digital Library, IEEE Xplore, CORE, PubMed, and Google Scholar databases to identify SUS scores related to the usability of DHAs for meta-analysis. This study included papers that published the SUS scores of the evaluated DHAs from 2011 to 2021 to get a 10-year representation. In total, 117 SUS scores for 114 DHAs were identified. R Studio and the R programming language were used to model the DHA SUS distribution, with a 1-sample, 2-tailed t test used to compare this distribution with the standard SUS distribution. RESULTS: The mean SUS score when all the collected apps were included was 76.64 (SD 15.12); however, this distribution exhibited asymmetrical skewness (-0.52) and was not normally distributed according to Shapiro-Wilk test (P=.002). The mean SUS score for "physical activity" apps was 83.28 (SD 12.39) and drove the skewness. Hence, the mean SUS score for all collected apps excluding "physical activity" apps was 68.05 (SD 14.05). A 1-sample, 2-tailed t test indicated that this health app SUS distribution was not statistically significantly different from the standard SUS distribution (P=.98). CONCLUSIONS: This study concludes that the SUS and the widely accepted benchmark of a mean SUS score of 68 (SD 12.5) are suitable for evaluating the usability of DHAs. We speculate as to why physical activity apps received higher SUS scores than expected. A template for reporting mean SUS scores to facilitate meta-analysis is proposed, together with future work that could be done to further examine the SUS benchmark scores for DHAs.

Exploring decision making 'noise' when interpreting the electrocardiogram in the context of cardiac cath lab activation.

In this commentary paper, we discuss the use of the electrocardiogram to help clinicians make diagnostic and patient referral decisions in acute care settings. The paper discusses the factors that are likely to contribute to the variability and noise in the clinical decision making process for catheterization lab activation. These factors include the variable competence in reading ECGs, the intra/inter rater reliability, the lack of standard ECG training, the various ECG machine and filter settings, cognitive biases (such as automation bias which is the tendency to agree with the computer-aided diagnosis or AI diagnosis), the order of the information being received, tiredness or decision fatigue as well as ECG artefacts such as the signal noise or lead misplacement. We also discuss potential research questions and tools that could be used to mitigate this 'noise' and improve the quality of ECG based decision making.

An insight into the current perceptions of UK radiographers on the future impact of AI on the profession: A cross-sectional survey.

INTRODUCTION: As a profession, radiographers have always been keen on adapting and integrating new technologies. The increasing integration of artificial intelligence (AI) into clinical practice in the last five years has been met with scepticism by some, who predict the demise of the profession, whilst others suggest a bright future with AI, full of opportunities and synergies. Post COVID-19 pandemic need for economic recovery and a backlog of medical imaging and reporting may accelerate the adoption of AI. It is therefore timely to appreciate practitioners' perceptions of AI used in clinical practice and their perception of the short-term impact on the profession. AIM: This study aims to explore the perceptions of AI in the UK radiography workforce and to investigate its current AI applications and future technological expectations of radiographers. METHODS: An online survey (QualtricsⓇ) was created by a team of radiography AI experts. The survey was disseminated via social media and professional networks in the UK. Demographic information and perceptions of the impact of AI on several aspects of the radiography profession were gathered, including the current use of AI in practice, future expectations and the perceived impact of AI on the profession. RESULTS: 411 responses were collected (80% diagnostic radiographers (DR); 20% therapeutic radiographers (TR)). Awareness of AI used in clinical practice is low, with DR respondents suggesting AI will have the most value/potential in cross sectional imaging and image reporting. TR responses linked AI as having most value in treatment planning, contouring, and image acquisition/matching. Respondents felt that AI will impact radiographers' daily work (DR, 79.6%; TR, 88.9%) by standardising some aspects of patient care and technical factors of radiography practice. A mixed response about impact on careers was reported. CONCLUSIONS: Respondents were unsure about the ways in which AI is currently used in practice and how AI will impact on careers in the future. It was felt that AI integration will lead to increased job opportunities to contribute to decision making as an end user. Job security was not identified as a cause for concern.

WaSP-ECG: A Wave Segmentation Pretraining Toolkit for Electrocardiogram Analysis.

Introduction: Representation learning allows artificial intelligence (AI) models to learn useful features from large, unlabelled datasets. This can reduce the need for labelled data across a range of downstream tasks. It was hypothesised that wave segmentation would be a useful form of electrocardiogram (ECG) representation learning. In addition to reducing labelled data requirements, segmentation masks may provide a mechanism for explainable AI. This study details the development and evaluation of a Wave Segmentation Pretraining (WaSP) application. Materials and Methods: Pretraining: A non-AI-based ECG signal and image simulator was developed to generate ECGs and wave segmentation masks. U-Net models were trained to segment waves from synthetic ECGs. Dataset: The raw sample files from the PTB-XL dataset were downloaded. Each ECG was also plotted into an image. Fine-tuning and evaluation: A hold-out approach was used with a 60:20:20 training/validation/test set split. The encoder portions of the U-Net models were fine-tuned to classify PTB-XL ECGs for two tasks: sinus rhythm (SR) vs atrial fibrillation (AF), and myocardial infarction (MI) vs normal ECGs. The fine-tuning was repeated without pretraining. Results were compared. Explainable AI: an example pipeline combining AI-derived segmentation masks and a rule-based AF detector was developed and evaluated. Results: WaSP consistently improved model performance on downstream tasks for both ECG signals and images. The difference between non-pretrained models and models pretrained for wave segmentation was particularly marked for ECG image analysis. A selection of segmentation masks are shown. An AF detection algorithm comprising both AI and rule-based components performed less well than end-to-end AI models but its outputs are proposed to be highly explainable. An example output is shown. Conclusion: WaSP using synthetic data and labels allows AI models to learn useful features for downstream ECG analysis with real-world data. Segmentation masks provide an intermediate output that may facilitate confidence calibration in the context of end-to-end AI. It is possible to combine AI-derived segmentation masks and rule-based diagnostic classifiers for explainable ECG analysis.

Remote and wearable ECG devices with diagnostic abilities in adults: A state-of-the-science scoping review.

The electrocardiogram (ECG) records the electrical activity in the heart in real time, providing an important opportunity to detecting various cardiac pathologies. The 12-lead ECG currently serves as the "standard" ECG acquisition technique for diagnostic purposes for many cardiac pathologies other than arrhythmias. However, the technical aspects of acquiring a 12-lead ECG are not easy. and its usage currently is restricted to trained medical personnel, which limits the scope of its usefulness. Remote and wearable ECG devices have attempted to bridge this gap by enabling patients to take their own ECG using a simplified method at the expense of a reduced number of leads, usually a single-lead ECG. In this review, we summarize the studies that investigated the use of remote ECG devices and their clinical utility in diagnosing cardiac pathologies. Eligible studies discussed Food and Drug Administration-cleared, commercially available devices that were validated in an adult population. We summarize technical logistics of signal quality and device reliability, dimensional and functional features, and diagnostic value. Our synthesis shows that reduced-set ECG wearables have huge potential for long-term monitoring, particularly if paired with real-time notification techniques. Such capabilities make them primarily useful for abnormal rhythm detection, and there is sufficient evidence that a remote ECG device can be superior to the traditional 12-lead ECG in diagnosing specific arrhythmias such as atrial fibrillation. However, this review identifies important challenges faced by this technology and highlights the limited availability of clinical research examining their usefulness.

Pediatric defibrillation shocks alone do not cause heart damage in a porcine model.

AIM: Automated external defibrillators (AEDs) use various shock protocols with different characteristics when deployed in pediatric mode. The aim of this study is to assess and compare the safety and efficacy of different AED pediatric protocols using novel experimental approaches. METHODS: Two defibrillation protocols (A and B) were assessed across two studies: Protocol A: escalating (50-75-90 J) defibrillation waveform with higher voltage, shorter duration and equal phase durations. Protocol B; non-escalating (50-50-50 J) defibrillation waveform with lower voltage, longer duration and unequal phase durations.Experiment 1: Isolated shock damage was assessed following shocks to 12 anesthetized pigs. Animals were randomized into two groups, receiving three shocks from Protocol A (50-75-90 J) or B (50-50-50 J). Cardiac function, cardiac troponin I (cTnI), creatine phosphokinase (CPK) and histopathology were analyzed. Experiment 2: Defibrillation safety and efficacy were assessed through shock success, ROSC, ST-segment deviation and contractility following 16 randomized shocks from protocol A or B delivered to 10 anesthetized pigs in VF. RESULTS: Experiment 1: No clinically meaningful difference in cTnI, CPK, ST-segment deviation, ejection fraction or histopathological damage was observed following defibrillation with either protocol. No difference was observed between protocols at any timepoint. Experiment 2: all defibrillation types demonstrated shock success and ROSC ≥ 97.5%. Post-ROSC contractility was similar between protocols. CONCLUSIONS: There is no evidence that administration of clinically relevant shock sequences, without experimental confounders, result in significant myocardial damage in this model of pediatric resuscitation. Typical variations in AED pediatric mode settings do not affect defibrillation safety and efficacy.

Using latent class analysis to identify clinical features of patients with occlusive myocardial infarction: Preangiogram prediction remains difficult.

BACKGROUND: Treatment decisions in myocardial infarction (MI) are currently stratified by ST elevation (ST-elevation myocardial infarction [STEMI]) or lack of ST elevation (non-ST elevation myocardial infarction [NSTEMI]) on the electrocardiogram. This arose from the assumption that ST elevation indicated acute coronary artery occlusion (OMI). However, one-quarter of all NSTEMI cases are an OMI, and have a higher mortality. The purpose of this study was to identify features that could help identify OMI. METHODS: Prospectively collected data from patients undergoing percutaneous coronary intervention (PCI) was analyzed. Data included presentation characteristics, comorbidities, treatments, and outcomes. Latent class analysis was undertaken, to determine patterns of presentation and history associated with OMI. RESULTS: A total of 1412 patients underwent PCI for acute MI, and 263 were diagnosed as OMI. Compared to nonocclusive MI, OMI patients are more likely to have fewer comorbidities but no difference in cerebrovascular disease and increased acute mortality (4.2% vs. 1.1%; p < .001). Of OMI, 29.5% had delays to their treatment such as immediate reperfusion therapy. With latent class analysis, while clusters of similar patients are observed in the data set, the data available did not usefully identify patients with OMI compared to non-OMI. CONCLUSION: Features between OMI and STEMI are broadly very similar. However, there was no difference in age and risk of cerebrovascular disease in the OMI/non-OMI group. There are no reliable characteristics therefore for identifying OMI versus non-OMI. Delays to treatment also suggest that OMI patients are still missing out on optimal treatment. An alternative strategy is required to improve the identification of OMI patients.

User experience of home-based AbC-19 SARS-CoV-2 antibody rapid lateral flow immunoassay test.

The urgent need to scale up testing capacity during the COVID-19 pandemic has prompted the rapid development of point-of-care diagnostic tools such as lateral flow immunoassays (LFIA) for large-scale community-based rapid testing. However, studies of how the general public perform when using LFIA tests in different environmental settings are scarce. This user experience (UX) study of 264 participants in Northern Ireland aimed to gather a better understanding of how self-administered LFIA tests were performed by the general public at home. The UX performance was assessed via analysis of a post-test questionnaire including 30 polar questions and 11 7-point Likert scale questions, which covers the multidimensional aspects of UX in terms of ease of use, effectiveness, efficiency, accuracy and satisfaction. Results show that 96.6% of participants completed the test with an overall average UX score of 95.27% [95% confidence interval (CI) 92.71-97.83%], which suggests a good degree of user experience and effectiveness. Efficiency was assessed based on the use of physical resources and human support received, together with the mental effort of self-administering the test measured via NASA Task Load Index (TLX). The results for six TLX subscales show that the participants scored the test highest for mental demand and lowest for physical demand, but the average TLX score suggests that the general public have a relatively low level of mental workload when using LFIA self-testing at home. Five printed LFIA testing results (i.e. the 'simulated' results) were used as the ground truth to assess the participant's performance in interpreting the test results. The overall agreement (accuracy) was 80.63% [95% CI 75.21-86.05%] with a Kappa score 0.67 [95% CI 0.58-0.75] indicating substantial agreement. The users scored lower in confidence when interpreting test results that were weak positive cases (due to the relatively low signal intensity in the test-line) compared to strong positive cases. The end-users also found that the kit was easier to use than they expected (p < 0.001) and 231 of 264 (87.5%) reported that the test kit would meet their requirements if they needed an antibody testing kit. The overall findings provide an insight into the opportunities for improving the design of self-administered SARS-CoV-2 antibody testing kits for the general public and to inform protocols for future UX studies of LFIA rapid test kits.

A clinician's guide to understanding and critically appraising machine learning studies: a checklist for Ruling Out Bias Using Standard Tools in Machine Learning (ROBUST-ML).

Developing functional machine learning (ML)-based models to address unmet clinical needs requires unique considerations for optimal clinical utility. Recent debates about the rigours, transparency, explainability, and reproducibility of ML models, terms which are defined in this article, have raised concerns about their clinical utility and suitability for integration in current evidence-based practice paradigms. This featured article focuses on increasing the literacy of ML among clinicians by providing them with the knowledge and tools needed to understand and critically appraise clinical studies focused on ML. A checklist is provided for evaluating the rigour and reproducibility of the four ML building blocks: data curation, feature engineering, model development, and clinical deployment. Checklists like this are important for quality assurance and to ensure that ML studies are rigourously and confidently reviewed by clinicians and are guided by domain knowledge of the setting in which the findings will be applied. Bridging the gap between clinicians, healthcare scientists, and ML engineers can address many shortcomings and pitfalls of ML-based solutions and their potential deployment at the bedside.

Upper Limb Position Tracking with a Single Inertial Sensor Using Dead Reckoning Method with Drift Correction Techniques.

Inertial sensors are widely used in human motion monitoring. Orientation and position are the two most widely used measurements for motion monitoring. Tracking with the use of multiple inertial sensors is based on kinematic modelling which achieves a good level of accuracy when biomechanical constraints are applied. More recently, there is growing interest in tracking motion with a single inertial sensor to simplify the measurement system. The dead reckoning method is commonly used for estimating position from inertial sensors. However, significant errors are generated after applying the dead reckoning method because of the presence of sensor offsets and drift. These errors limit the feasibility of monitoring upper limb motion via a single inertial sensing system. In this paper, error correction methods are evaluated to investigate the feasibility of using a single sensor to track the movement of one upper limb segment. These include zero velocity update, wavelet analysis and high-pass filtering. The experiments were carried out using the nine-hole peg test. The results show that zero velocity update is the most effective method to correct the drift from the dead reckoning-based position tracking. If this method is used, then the use of a single inertial sensor to track the movement of a single limb segment is feasible.

Beauty Is in the AI of the Beholder: Are We Ready for the Clinical Integration of Artificial Intelligence in Radiography? An Exploratory Analysis of Perceived AI Knowledge, Skills, Confidence, and Education Perspectives of UK Radiographers.

Introduction: The use of artificial intelligence (AI) in medical imaging and radiotherapy has been met with both scepticism and excitement. However, clinical integration of AI is already well-underway. Many authors have recently reported on the AI knowledge and perceptions of radiologists/medical staff and students however there is a paucity of information regarding radiographers. Published literature agrees that AI is likely to have significant impact on radiology practice. As radiographers are at the forefront of radiology service delivery, an awareness of the current level of their perceived knowledge, skills, and confidence in AI is essential to identify any educational needs necessary for successful adoption into practice. Aim: The aim of this survey was to determine the perceived knowledge, skills, and confidence in AI amongst UK radiographers and highlight priorities for educational provisions to support a digital healthcare ecosystem. Methods: A survey was created on Qualtrics® and promoted via social media (Twitter®/LinkedIn®). This survey was open to all UK radiographers, including students and retired radiographers. Participants were recruited by convenience, snowball sampling. Demographic information was gathered as well as data on the perceived, self-reported, knowledge, skills, and confidence in AI of respondents. Insight into what the participants understand by the term "AI" was gained by means of a free text response. Quantitative analysis was performed using SPSS® and qualitative thematic analysis was performed on NVivo®. Results: Four hundred and eleven responses were collected (80% from diagnostic radiography and 20% from a radiotherapy background), broadly representative of the workforce distribution in the UK. Although many respondents stated that they understood the concept of AI in general (78.7% for diagnostic and 52.1% for therapeutic radiography respondents, respectively) there was a notable lack of sufficient knowledge of AI principles, understanding of AI terminology, skills, and confidence in the use of AI technology. Many participants, 57% of diagnostic and 49% radiotherapy respondents, do not feel adequately trained to implement AI in the clinical setting. Furthermore 52% and 64%, respectively, said they have not developed any skill in AI whilst 62% and 55%, respectively, stated that there is not enough AI training for radiographers. The majority of the respondents indicate that there is an urgent need for further education (77.4% of diagnostic and 73.9% of therapeutic radiographers feeling they have not had adequate training in AI), with many respondents stating that they had to educate themselves to gain some basic AI skills. Notable correlations between confidence in working with AI and gender, age, and highest qualification were reported. Conclusion: Knowledge of AI terminology, principles, and applications by healthcare practitioners is necessary for adoption and integration of AI applications. The results of this survey highlight the perceived lack of knowledge, skills, and confidence for radiographers in applying AI solutions but also underline the need for formalised education on AI to prepare the current and prospective workforce for the upcoming clinical integration of AI in healthcare, to safely and efficiently navigate a digital future. Focus should be given on different needs of learners depending on age, gender, and highest qualification to ensure optimal integration.

Machine learning with electrocardiograms: A call for guidelines and best practices for 'stress testing' algorithms.

This paper provides a brief description of how computer programs are used to automatically interpret electrocardiograms (ECGs), and also provides a discussion regarding new opportunities. The algorithms that are typically used today in hospitals are knowledge engineered where a computer programmer manually writes computer code and logical statements which are then used to deduce a possible diagnosis. The computer programmer's code represents the criteria and knowledge that is used by clinicians when reading ECGs. This is in contrast to supervised machine learning (ML) approaches which use large, labelled ECG datasets to induct their own 'rules' to automatically classify ECGs. Although there are many ML techniques, deep neural networks are being increasingly explored as ECG classification algorithms when trained on large ECG datasets. Whilst this paper presents some of the pros and cons of each of these approaches, perhaps there are opportunities to develop hybridised algorithms that combine both knowledge and data driven techniques. In this paper, it is pointed out that open ECG data can dramatically influence what international ECG ML researchers focus on and that, ideally, open datasets could align with real world clinical challenges. In addition, some of the pitfalls and opportunities for ML with ECGs are outlined. A potential opportunity for the ECG community is to provide guidelines to researchers to help guide ECG ML practices. For example, whilst general ML guidelines exist, there is perhaps a need to recommend approaches for 'stress testing' and evaluating ML algorithms for ECG analysis, e.g. testing the algorithm with noisy ECGs and ECGs acquired using common lead and electrode misplacements. This paper provides a primer on ECG ML and discusses some of the key challenges and opportunities.

User experience analysis of AbC-19 Rapid Test via lateral flow immunoassays for self-administrated SARS-CoV-2 antibody testing.

Lateral flow immunoassays are low cost, rapid and highly efficacious point-of-care devices, which have been used for SARS-CoV-2 antibody testing by professionals. However, there is a lack of understanding about how self-administered tests are used by the general public for mass testing in different environmental settings. The purpose of this study was to assess the user experience (UX) (including usability) of a self-testing kit to identify COVID-19 antibodies used by a representative sample of the public in their cars, which included 1544 participants in Northern Ireland. The results based on 5-point Likert ratings from a post-test questionnaire achieved an average UX score of 96.03% [95% confidence interval (CI) 95.05-97.01%], suggesting a good degree of user experience. The results of the Wilcoxon rank sum tests suggest that UX scores were independent of the user's age and education level although the confidence in this conclusion could be strengthened by including more participants aged younger than 18 and those with only primary or secondary education. The agreement between the test result as interpreted by the participant and the researcher was 95.85% [95% CI 94.85-96.85%], Kappa score 0.75 [95% CI 0.69-0.81] (indicating substantial agreement). Text analysis via the latent Dirichlet allocation model for the free text responses in the survey suggest that the user experience could be improved for blood-sample collection, by modifying the method of sample transfer to the test device and giving clearer instructions on how to interpret the test results. The overall findings provide an insight into the opportunities for improving the design of SARS-CoV-2 antibody testing kits to be used by the general public and therefore inform protocols for future user experience studies of point-of-care tests.

Digital reminiscence app co-created by people living with dementia and carers: Usability and eye gaze analysis.

BACKGROUND: This research reports on a pilot study that examined the usability of a reminiscence app called 'InspireD' using eye tracking technology. The InspireD app is a bespoke digital intervention aimed at supporting personalized reminiscence for people living with dementia and their carers. The app was developed and refined in two co-creation workshops and subsequently tested in a third workshop using eye tracking technology. INTERVENTION: Eye tracking was used to gain insight into the user's cognition since our previous work showed that the think-aloud protocol can add to cognitive burden for people living with dementia while also making the test more unnatural. RESULTS: Results showed that there were no barriers to using a wearable eye tracker in this setting and participants were able to use the reminiscence app freely. However, some tasks required prompts from the observer when difficulties arose. While prompts are not normally used in usability testing (as some argue the prompting defeats the purpose of testing), we used 'prompt frequency' as a proxy for measuring the intuitiveness of the task. There was a correlation between task completion rates and prompt frequency. Results also showed that people living with dementia had fewer gaze fixations when compared to their carers. Carers had greater fixation and saccadic frequencies when compared to people living with dementia. This perhaps indicates that people living with dementia take more time to scan and consume information on an app. A number of identified usability issues are also discussed in the paper. PATIENT OR PUBLIC CONTRIBUTION: The study presents findings from three workshops which looked at user needs analysis, feedback and an eye tracking usability test combined involving 14 participants, 9 of whom were people living with dementia and the remaining 5 were carers.

Comparing Single-Page, Multipage, and Conversational Digital Forms in Health Care: Usability Study.

BACKGROUND: Even in the era of digital technology, several hospitals still rely on paper-based forms for data entry for patient admission, triage, drug prescriptions, and procedures. Paper-based forms can be quick and convenient to complete but often at the expense of data quality, completeness, sustainability, and automated data analytics. Digital forms can improve data quality by assisting the user when deciding on the appropriate response to certain data inputs (eg, classifying symptoms). Greater data quality via digital form completion not only helps with auditing, service improvement, and patient record keeping but also helps with novel data science and machine learning research. Although digital forms are becoming more prevalent in health care, there is a lack of empirical best practices and guidelines for their design. The study-based hospital had a definite plan to abolish the paper form; hence, it was not necessary to compare the digital forms with the paper form. OBJECTIVE: This study aims to assess the usability of three different interactive forms: a single-page digital form (in which all data input is required on one web page), a multipage digital form, and a conversational digital form (a chatbot). METHODS: The three digital forms were developed as candidates to replace the current paper-based form used to record patient referrals to an interventional cardiology department (Cath-Lab) at Altnagelvin Hospital. We recorded usability data in a counterbalanced usability test (60 usability tests: 20 subjects×3 form usability tests). The usability data included task completion times, System Usability Scale (SUS) scores, User Experience Questionnaire data, and data from a postexperiment questionnaire. RESULTS: We found that the single-page form outperformed the other two digital forms in almost all usability metrics. The mean SUS score for the single-page form was 76 (SD 15.8; P=.01) when compared with the multipage form, which had a mean score of 67 (SD 17), and the conversational form attained the lowest scores in usability testing and was the least preferred choice of users, with a mean score of 57 (SD 24). An SUS score of >68 was considered above average. The single-page form achieved the least task completion time compared with the other two digital form styles. CONCLUSIONS: In conclusion, the digital single-page form outperformed the other two forms in almost all usability metrics; it had the least task completion time compared with those of the other two digital forms. Moreover, on answering the open-ended question from the final customized postexperiment questionnaire, the single-page form was the preferred choice.