Artificial Intelligence Algorithms Research Articles

Integrated water resource management in the Segura Hydrographic Basin: An artificial intelligence approach

Managing resources effectively in uncertain demand, variable availability, and complex governance policies is a significant challenge. This paper presents a paradigmatic framework for addressing these issues in water management scenarios by integrating advanced physical modelling, remote sensing techniques, and Artificial Intelligence algorithms. The proposed approach accurately predicts water availability, estimates demand, and optimizes resource allocation on both short- and long-term basis, combining a comprehensive hydrological model, agronomic crop models for precise demand estimation, and Mixed-Integer Linear Programming for efficient resource distribution. In the study case of the Segura Hydrographic Basin, the approach successfully allocated approximately 642 million cubic meters (hm3) of water over six months, minimizing the deficit to 9.7% of the total estimated demand. The methodology demonstrated significant environmental benefits, reducing CO2 emissions while optimizing resource distribution. This robust solution supports informed decision-making processes, ensuring sustainable water management across diverse contexts. The generalizability of this approach allows its adaptation to other basins, contributing to improved governance and policy implementation on a broader scale. Ultimately, the methodology has been validated and integrated into the operational water management practices in the Segura Hydrographic Basin in Spain.

Capabilities of cellebrite universal forensics extraction device in mobile device forensics

The powerful digital forensics tool cellebrite universal forensics extraction device (UFED) extracts and analyzes mobile device data, helping investigators solve criminal and cybersecurity cases. Advanced methods and algorithms allow Cellebrite UFED to recover data from erased or obscured devices. Cellebrite UFED can pull data from call logs, texts, emails, and social media, providing valuable evidence for investigations. The use of smartphones and tablets in personal and professional settings has spurred the development of mobile device forensics. The intuitive user interface speeds up data extraction and analysis, revealing crucial information. It can decrypt encrypted data, recover deleted files, and extract data from multiple devices. The sector's best data extraction functionality, Cellebrite UFED, helps forensic analysts gather crucial evidence for investigations. Legal and ethical considerations are crucial in mobile device forensics. Legal considerations include allowing access to data, protecting privacy, and adhering to chain of custody protocols. Ethics include transparency, defamation, and information exploitation protection. Using Cellebrite UFED, researchers can navigate complex data on mobile devices more efficiently and precisely. Artificial intelligence (AI) and machine learning (ML) algorithms may automate data extraction in future tools. Examiners must train, maintain, and establish clear protocols for using Cellebrite UFED in forensic investigations.

Breast cancer survival prediction using an automated mitosis detection pipeline.

Mitotic count (MC) is the most common measure to assess tumor proliferation in breast cancer patients and is highly predictive of patient outcomes. It is, however, subject to inter- and intraobserver variation and reproducibility challenges that may hamper its clinical utility. In past studies, artificial intelligence (AI)-supported MC has been shown to correlate well with traditional MC on glass slides. Considering the potential of AI to improve reproducibility of MC between pathologists, we undertook the next validation step by evaluating the prognostic value of a fully automatic method to detect and count mitoses on whole slide images using a deep learning model. The model was developed in the context of the Mitosis Domain Generalization Challenge 2021 (MIDOG21) grand challenge and was expanded by a novel automatic area selector method to find the optimal mitotic hotspot and calculate the MC per 2 mm2. We employed this method on a breast cancer cohort with long-term follow-up from the University Medical Centre Utrecht (N = 912) and compared predictive values for overall survival of AI-based MC and light-microscopic MC, previously assessed during routine diagnostics. The MIDOG21 model was prognostically comparable to the original MC from the pathology report in uni- and multivariate survival analysis. In conclusion, a fully automated MC AI algorithm was validated in a large cohort of breast cancer with regard to retained prognostic value compared with traditional light-microscopic MC.

DocBot AI as Your Personal Health Assistant: Bridging the Gap in Consultations Revolutionizing Healthcare Using Natural Language Processing and Artificial Intelligence

The DocBot Chat Bot/AI system is a novel solution to the challenge of acquiring accurate and personalized health information. By utilizing cutting-edge artificial intelligence (AI) algorithms and natural language processing (NLP) methods, the DocBot system creates an engaging and intuitive interface that lets people convey their health concerns and get personalized advice. This study provides an in-depth assessment of the DocBot system, stressing its key features, AI algorithms, advantages, difficulties, and potential applications. To improve healthcare communication and encourage well-informed decision-making, it is important to highlight DocBot's transformative potential in providing users with accurate and trustworthy health knowledge.

A Review of Advancement in using Generative AI for Detecting Skin Cancer

Abstract: Skin cancer is a very common type of cancer that is asymptomatic in the early stage of development and is difficult to diagnose and treat. More and more people have used generative AI methods to enhance the speed and accuracy of skin cancer prediction in the recent past. Current techniques of artificial intelligence and deep learning algorithms are performing quite effectively in the field of dermoscopy and identification of malignant lesions. This study overviews the recent developments in Skin Cancer Detection Approaches based on emerging AI methods like Generative Adversarial Networks (GAN); Variational AutoEncoders (VAE); and derivatives of both. In this study, we discussed several approaches to implementing generative AI for skin cancer detection, the advantages and disadvantages of each approach, and identified research that has revealed potential in the field

Harnessing AI for Network Security and DDoS Attack Detection

This paper investigates the utilization of artificial intelligence (AI) techniques for network security and detection of Distributed Denial of Service (DDoS) attacks. It explores the role of AI algorithms, machine learning models, and anomaly detection techniques in enhancing network security, mitigating cyber threats, and improving incident response in the face of DDoS attacks. The paper discusses real-world implementations and case studies showcasing the effectiveness of AI in network security and DDoS attack detection.

Exploring the synergy between generative AI and software engineering: Automating code optimization and bug fixing

As applied to software engineering, generative AI is quickly transitioning from a zero-sum industry game changer into the primary automation tool for code optimization, bug identification, and problem-solving. This technology takes advantage of artificial intelligence algorithms within machine learning models to analyze and write code, resulting in improved quality and speed of an application development process. The generative AI replenishes productivity in development work and enhances centralization between development work teams through code handling and intelligent suggestions for essential codes. However, the integration of AI in software engineering poses the following problems and ethical questions: the question of accuracy, bias, and data. This paper will review the existing knowledge on generative AI in software engineering regarding its current use, future evolutions and advancements, issues and limitations, and ethical factors in using this technology. This paper considers these aspects to give a global outlook on how generative AI will transform software development in the future and how responsible AI should be employed.

Social sustainability: Viability analysis of social firms

AbstractSocial enterprises are firms that actively create jobs for people with disabilities and reduce social and labour inequalities, thereby participating in the social economy. Due to their importance to society and the work they do towards diversity integration and social sustainability, they have recently attracted much attention in academic literature. For this reason, the main objective of this study was to analyse the survival of social enterprises, identifying key variables that condition their continuity in the market or their failure. The initial sample consisted of 999 social enterprises for the year 2022. The Altman Z‐score and artificial intelligence (AI) algorithms were used to obtain the basic survival patterns. The main findings were that, on average, social enterprises are highly experienced companies and only one‐third of them are at risk of bankruptcy. This means that most of these enterprises can continue their social function. Moreover, the results of return on assets (ROA), equity and debt turnover can predict being at risk of bankruptcy of social enterprises. This study contributes to the scarce academic literature on social enterprises and promotes the existence of such enterprises for social and economic sustainability.

Artificial intelligence for automated left ventricular systolic dysfunction detection using a wearable cardiac patch incorporated with synchronized phonocardiogram and electrocardiogram

Abstract Background Left ventricular systolic dysfunction (LVSD) is characterized by a reduced left ventricular ejection fraction (LVEF) and is associated with three times the risk of developing Heart failure (HF). We aimed to test an artificial intelligence (AI) algorithm applied to synchronized phonocardiography (PCG) and electrocardiography (ECG) signals, recorded with a wearable device, to validate its potential as a screening tool for LVSD. Methods Using ECG and PCG data collected from 1960 admitted patients in a hospital, we trained an AI algorithm to detect the LVSD. There are two ways of defining LVSD, one group is defined by LVEF <50% and the other is defined by LVEF 40% as measured by echocardiography. The AI algorithm followed a two-step detection structure. In the first phase, ECG signals were processed by wavelet denoising and baseline wander correction, and PCG signals were converted to the frequency domain via wavelet transformation. We developed a 1D CNN-based U-Net variant model to pinpoint electromechanical activation time (EMAT). Addressing the challenge of processing multimodal signals, we devised a contrastive learning approach that mandates the encoders for different modalities to generate embeddings within the same feature space. This trained encoder, along with the decoder in U-Net, is subsequently fine-tuned for EMAT detection. The AI-EMAT% was obtained by adjusting detected EMAT based on RR interval. For the second phase, random forest regression is used to estimate LVEF, incorporating features such as AI-EMAT% and clinical data, including age and gender. For LVSD detection, we evaluated the performance of methods by bifurcating AI-estimated LVEF% (AI-predict LVEF%) based on thresholds of 40% and 50%, and by directly estimating AI-EMAT% cutoff. The evaluation of LVSD detection algorithm, conducted via 2-fold cross-validation, involves dividing the dataset into two subsets, with each alternately serving as training and testing sets to ensure a fair assessment. Results We recruited 1960 patients (1430 [72.96%] male). 357 (18.21%) had an ejection fraction of 50% or lower, and 157 (8.01%) had an ejection fraction of 40% or lower. In detecting LVSD of LVEF<50%, AI-EMAT% yielded an AUROC of 0.86, sensitivity of 73.7%, and specificity of 80.7%. While AI-predict LVEF% approach resulted in an AUROC of 0.89, sensitivity of 81.8%, and specificity of 81.5%. In detecting LVSD of LVEF<40%, AI-EMAT% yielded an AUROC of 0.89, sensitivity of 85.4%, and specificity of 76.0%. While AI-predict LVEF% outputs resulted in an AUROC of 0.91, sensitivity of 92.4%, and specificity of 76.8%. Conclusions We developed an automated algorithm to detect LVSD using a wearable device that incorporated synchronized PCG and ECG signals, demonstrating robust performance across different subgroups. This approach represents a strategy for automated screening of LV systolic dysfunction, particularly beneficial in resource-limited settings.Baseline and AUROCAI algorithm

Multi-modal artificial intelligence-based prediction for cardiovascular mortality after transcatheter aortic valve implantation: a time-to-event survival prediction

Abstract Background Predicting long-term outcomes following Transcatheter Aortic Valve Implantation (TAVI) presents a significant challenge. Objective We aimed to predict post-TAVI long-term cardiovascular mortality using multimodal data and employing time-to-event artificial intelligence (AI) algorithms. Methods This study enrolled 3,973 consecutive patients from the prospective TAVI registry (between August 2007 and June 2023). Multimodal information was collected and extracted as numerical data for each patient, encompassing clinical history, medication, laboratory, procedural characteristics, ECG, angiography, echocardiography (TTE and TEE), and CT. Clinical outcomes were adjudicated by an independent clinical event committee. From the multimodal information, 147 features were extracted at baseline before the TAVI procedure. The datasets were divided into a training/validation set (80%) and a hold-out test set (20%) for model development and evaluation, respectively. Cox-LASSO feature selection was employed to identify the most relevant features, followed by the implementation of Random Survival Forest (RSF), Gradient Boosting Survival Analysis (GBSA), and Survival Support Vector Machine (SSVM) to predict the timing and occurrence of cardiac death events in patients. Feature selection, model parameters, and hyperparameters selection were performed using 10-fold cross-validation on the training set. Model performance was evaluated using different metrics, such as the concordance index (c-index) and cumulative Area Under the Curve (C-AUC) on a 20% untouched test set. Results In the patient cohort, 28.4% experienced cardiovascular death. Features were selected from various modalities, including 22 from clinical history, two from medication, eight from laboratory tests, one from ECG, eight from CT scans, and five from echocardiography and angiography. In the univariate Cox-PH model, all features exhibited a C-index ranging from 0.50 to 0.57. In the test set, the conventional Multivariate Cox-PH model achieves a c-index of 0.59 (C-AUC: 0.56). RSF, GBSA, and SSVM achieve a C-index of 0.70 (C-AUC: 0.69), 0.66 (C-AUC: 0.64), and 0.69 (C-AUC: 0.68), respectively, in the holdout test set. Conclusion Integrating multimodal information, encompassing imaging, signal, laboratory data, and clinical history, with AI algorithms has markedly enhanced the prediction of long-term cardiac outcomes following the TAVI procedure. The initially developed model can predict the probability of cardiac death up to 10 years. Furthermore, the RFS model significantly outperformed the conventional Cox model in predicting time-to-event outcomes.

AI-driven ethical issues in the development of personalised therapy for pediatric SIRS patients

Abstract Background Systemic Inflammatory Response Syndrome (SIRS) is a rare disease often presenting emergency. While the aetiology varies, current diagnostic criteria do not allow differentiation. Hopes are placed on OMICs research and development of Artificial intelligence (AI) algorithms for personalised disease management. Our report aims at identification and analysis of AI-driven ethical issues within the bigger ERA PerMed funded project of development of tailored immunotherapy for paediatric SIRS patients (TIPS) coordinated by Prof. Catharina Schütz from Technic University Dresden. Methods Systematic literature review followed by qualitative research methodology. Ten experts with different backgrounds were interviewed. Subsequent thematic analysis was performed by two independent researchers. Results Experts were on the opinion that AI can lead to a depart from individual approach in the medicine, as the physician may be induced to only act based on algorithms that are more or less representative (N = 2). Interviewees were aware that AI can lead to changes in the assessment of reality, as it is dependent on data quality (N = 4). AI-aided decision-making might be useful for special therapeutic situations, although it may not always capture the subtleties of individual situations (N = 2). Positive implications of AI were significantly more emphasised, though current frameworks were perceived as more limiting than enabling. Conclusions Personalised therapy is making its first steps while the ethical apprehensions has already been there with a strong emphasis on precautionary principle and resulting limitations on the application of AI-algorithms. The multifaceted nature of the AI-driven ethical issues in personalised medicine research requires an interdisciplinary discourse, strengthened ethics training of all professionals involved in the technology development and more active involvement of physicians in the shaping of AI. Key messages • AI algorithms promise therapeutic benefits, not achievable by other means, for specific categories of vulnerable pediatrics patients. • The multifaceted nature of the AI-driven ethical issues in personalised medicine research requires an interdisciplinary discourse and further ethics training of all stakeholders.

Attention as the dimension that deteriorates first in patients with cognitive impairment after myocardial infarction

Abstract Background Previous publications by our team indicated a high prevalence (approximately 35%) of cognitive impairment (CI) in patients after myocardial infarction (MI). Moreover, we observed that the presence of peri-infarction deficits was not always associated with their occurrence after 6 months of follow-up and vice versa. This was the basis for distinguishing 4 groups of patients: a) group 1 – CI present peri-infarction and after 6 months; b) group 2 – CI present only peri-infarction; c) group 3 – CI present only after 6 months; d) group 4 – no CI during the study. Purpose The aim of the current study is to analyse the type of CI in patients after MI with a focus on its changes within the 4 groups mentioned. Methods The study included 326 patients hospitalised for MI treated by percutaneous coronary intervention. Cognitive function was assessed twice using the Mini–Mental State Examination (MMSE) and the Clock Drawing Test (CDT)-Shulman version during the first hospitalisation (on day 2-3 after MI) and 6 months later. In addition, the occurrence of sleep, aggression and mood disorders was assessed. The constructed research model was analysed with artificial intelligence algorithms. Results Persistent CI (group1) was found in 8.9%, peri-infarct CI (group 2) in 16.3%, and only after 6 months (group 3) in 7.7% of study participants (as assessed by MMSE). Analysis of the cognitive domains showed that in those who improve (group 2) or worsen cognitive function (group 3), analogous changes in attention function occurred, respectively improving in 81% and worsening in 100% of subjects in a given group. It was observed that there is a group of patients with a CI on the MMSE who perform CDT correctly, 12% peri-infarct and 11% at 6-month follow-up, respectively. Patients with a normal CDT score but CIs found in the MMSE have impaired attentional function regardless of the time of assessment. Conclusion CIs can occur both immediately after MI and during subsequent follow-up. However, after 6 months of intensive anti-atherosclerotic treatment (post-MI), cognitive functions improve in some patients, and deteriorate in others. At the same time, the cognitive function that undergoes similar changes in these groups is attention. The uniform type of impaired cognitive function and the lack of specific differences between the 4 groups allow us to assume a uniform etiology of cognitive deficits. Patients with normal CDT and disorders present in MMSE show deficits in attention functions. Performing CDT and using the MMSE component assessing attention function could prove sufficient for the initial assessment of cognitive functions in the population of patients after MI. The results require confirmation in a larger patient population.

Deep learning improves detection of aortic stenosis in b-mode ultrasound: results of a pivotal multicenter validation study using artificial intelligence

Abstract Background Aortic stenosis (AS) is a highly prevalent and morbid condition, with moderate or severe AS present in 9% of patients over 75 years. Despite improved diagnostic and therapeutic methods, AS remains underdiagnosed, particularly in Black patients and underserved communities. Echocardiography (echo) is effective for detecting AS but requires expertise to quantify and interpret valve gradients and valve area. We hypothesized that a deep learning (DL) algorithm could be trained to detect AS severity using B-mode (non-Doppler) echo alone, the modality most likely to be available in point-of-care ultrasound (POCUS) studies. Here, we present the results of a multicenter pivotal trial to test whether artificial intelligence (AI) can help increase the detection of AS from B-mode echoes. Methods A convolutional neural network (CNN) was trained to classify the severity of aortic stenosis from 3 B-mode echo views of the aortic valve (parasternal long- and short-axis and apical 5-chamber). The training dataset had over 500,000 cine loops from 29,527 echoes in 26,981 patients. A fully crossed multireader multicase study was conducted using 220 echoes never seen during algorithm development. 5 level II-trained cardiologists (referred to as POCUS readers as they only reviewed B-mode clips) were presented with these echoes twice at least a month apart, once with and again without AI interpretation of AS severity. Readers identified 2 dichotomous severity levels (either "suggestive of moderate to severe AS" or not), with ground truth established by 3 level III-trained cardiology echo experts. While experts had access to full studies (including Doppler) to adjudicate severity, POCUS readers had no Doppler or measurements of any kind available. Sensitivity and specificity were compared for the 5 POCUS readers when aided by the AI algorithm vs unaided, with similar tests of inter-reader agreement. Results There was a statistically significant improvement in sensitivity (+5.5%, 95% CI [1.5%, 9.5%], p=0.007) with stable specificity (-0.3%, 95% CI [-3.6%, 3.0%], p=NS) when the 5 POCUS readers received the assistance of the AI software. Fig 1 shows 2,500 bootstrapping assessments to better show consistency of these findings. In addition, the POCUS inter-reader agreement (overall percent agreement) improved by approximately 7% (p<0.001), becoming comparable to that of expert readers (Fig 2). Conclusions In a comprehensive multireader pivotal trial, an AI algorithm improved the performance of POCUS cardiologist readers to identify significant AS from limited B-mode echoes as might be obtained in screening or outreach settings. Sensitivity for detection of moderate-severe AS increased significantly by 5.5% with stable specificity. Inter-reader agreement improved to resemble that of expert cardiologists given access to full study information, including Doppler. AI may have a role in assisting detection of significant AS from limited B-mode echoes.Fig 1Fig 2

Barriers and facilitators for implementation of AI algorithms for ECG analysis and triage in patients with chest pain

Abstract Introduction Numerous artificial intelligence (AI) algorithms have been developed to assist in clinical decision-making, with a focus on areas like electrocardiogram (ECG) analysis due to the signal complexity and the varying levels of expertise among healthcare professionals. However, ensuring acceptance of these algorithms by end-users is paramount for successful implementation. Purpose This study aimed to identify barriers and facilitators related to implementation of an AI algorithm for ECG analysis and triage in patients with chest pain. Methods A three-round modified electronic Delphi study was conducted among Dutch potential end-users of an ECG-AI algorithm, including physicians, nurses and ambulance professionals. During the first round, participants brainstormed on barriers and facilitators, which were subsequently mapped to the Consolidated Framework for Implementation Research (CFIR). In the following two rounds, participants reviewed the identified barriers and facilitators for relevance and ranked them according to their importance. Results Twenty-five out of the initial forty respondents completed all three rounds, resulting in a dropout rate of 38%. Four barriers and twelve facilitators were identified. The most important facilitator was "Faster recognition of subtle ECG abnormalities", while "Poor model performance" emerged as the most important barrier. The most frequently mentioned corresponding CFIR domains were the innovation and inner setting domains. Conclusion The identification of barriers and facilitators provides valuable insights into the acceptance and utilisation of ECG-AI algorithms by end-users. By considering the CFIR domains, including innovation and inner setting, this study offers a comprehensive understanding of the contextual factors that influence the successful implementation of AI algorithms in healthcare settings. Addressing these barriers and facilitators will be crucial to facilitating future implementation processes.Barriers and facilitators

Rhythm classifier: artificial intelligence for rhythm classification in insertable cardiac monitor

Abstract Background Though detections improved significantly, separating atrial tachycardia/atrial flutter (AT/AFL) from atrial fibrillations (AF) in insertable cardiac monitors (ICM) remains a challenge. Objective Using artificial intelligence (AI) algorithms to classify AF and AT/AFL in ICM detected episodes, which may reduce episode review burden. Methods Both raw EGM signal and the QRS diminished version of the raw signal were used to extract 14 AF and AT/AFL features from each episode. Extracted features were then represented as an ensemble of features and used to train and test the AI model. Modified ResNet18 network was used as the AI model. All episodes from a random selection of 80% of patients were used as training dataset, another 10% were used as validation, and 10% were used as test dataset, to evaluate model performance. AI model probability threshold was chosen based on validation set results. Episode detection sensitivity, specificity, and area under the curve (AUC) of receiver operating characteristics (ROC) curve in the independent test dataset are reported. Results AF and AT/AFL episodes detected in 918 LINQ ICM patients (5,253 AF and 4,705 AT/AFL – 745 pts and 8,950 false detection episodes) were included. Training dataset consisted of 15,183 episodes (735 pts), validation set consisted of 1,901 episodes (91 pts), and test set consisted of 1,824 episodes (92 pts). At chosen threshold in validation set, AF and AT/AFL detection had sensitivity of 90% and 73.3%, respectively, and specificity of 93.2% and 95%, respectively, with AUC of 0.966. In test data, AF and AT/AFL obtained relative sensitivity of 84.7% and 85.7%, respectively, with specificity of 88% and 86.5%, respectively (Fig. A). With ROC-AUC of 0.956 (Fig. B), the relative diagnostic yield of AF and AT/AFL on test data was 95.6%. Conclusion The trained AI model provides a relative sensitivity of about 85% for AF vs. AT/AFL classification and reduces inappropriate ICM detections by 79%. Adding more episodes may improve the performance.

Novel Approaches for the Early Detection of Glaucoma Using Artificial Intelligence

Background: If left untreated, glaucoma—the second most common cause of blindness worldwide—causes irreversible visual loss due to a gradual neurodegeneration of the retinal ganglion cells. Conventional techniques for identifying glaucoma, like optical coherence tomography (OCT) and visual field exams, are frequently laborious and dependent on subjective interpretation. Through the fast and accurate analysis of massive amounts of imaging data, artificial intelligence (AI), in particular machine learning (ML) and deep learning (DL), has emerged as a promising method to improve the early detection and management of glaucoma. Aims: The purpose of this study is to examine the current uses of AI in the early diagnosis, treatment, and detection of glaucoma while highlighting the advantages and drawbacks of different AI models and algorithms. In addition, it aims to determine how AI technologies might transform glaucoma treatment and suggest future lines of inquiry for this area of study. Methods: A thorough search of databases, including Web of Science, PubMed, and Scopus, was carried out to find pertinent papers released until August 2024. The inclusion criteria were limited to research published in English in peer-reviewed publications that used AI, ML, or DL to diagnose or treat glaucoma in human subjects. Articles were chosen and vetted according to their quality, contribution to the field, and relevancy. Results: Convolutional neural networks (CNNs) and other deep learning algorithms are among the AI models included in this paper that have been shown to have excellent sensitivity and specificity in identifying glaucomatous alterations in fundus photos, OCT scans, and visual field tests. By automating standard screening procedures, these models have demonstrated promise in distinguishing between glaucomatous and healthy eyes, forecasting the course of the disease, and possibly lessening the workload of physicians. Nonetheless, several significant obstacles remain, such as the requirement for various training datasets, outside validation, decision-making transparency, and handling moral and legal issues. Conclusions: Artificial intelligence (AI) holds great promise for improving the diagnosis and treatment of glaucoma by facilitating prompt and precise interpretation of imaging data and assisting in clinical decision making. To guarantee wider accessibility and better patient results, future research should create strong generalizable AI models validated in various populations, address ethical and legal matters, and incorporate AI into clinical practice.

Performance of an automated echocardiographic artificial intelligence model. to detect subclinical heart failure with preserved ejection fraction (HFpEF) in community-dwelling older adults

Abstract Background Heart failure (HF) with preserved Ejection Fraction (HFpEF) is common among older adults and associated with a high burden of morbidity and mortality. Early identification of subclinical HFpEF, defined by abnormalities in cardiac structure and function without symptoms of HF, can prompt early initiation of evidence-based therapies to prevent HFpEF. However, scalable strategies for the identification of subclinical HFpEF are lacking. Purpose To evaluate the performance of a validated, automated echocardiographic artificial Intelligence algorithm (Ultromics, UK) to identify subclinical HFpEF among community-dwelling individuals without HF. Methods The study included participants of the Dallas Hearts and Minds Study without HF who underwent cardiovascular phenotyping with resting and exercise echocardiography and maximal cardiopulmonary exercise testing. The resting apical 4C echocardiographic images were analyzed by the AI echo algorithm to identify the HFpEF phenotype (AI-HFpEF). Subclinical HFpEF was defined by E/e’ >14 and Peak exercise oxygen uptake (VO2peak) <25th percentile for the cohort. The performance of the AI algorithm to detect subclinical HFpEF was determined using the area under the receiver operating curve and decision curve analysis and was compared with the previously validated H2FpEF score. The association between the AI-HFpEF phenotype and the presence of subclinical HFpEF, VO2peak, exercise E/e’, left ventricular strain, and left atrial strain was assessed using multivariable-adjusted logistic and linear regression models. Results The study included 511 participants (mean age: 61 y, 57% women, mean VO2peak = 16.9 ml/kg/min). Using the AI echo algorithm, subclinical HFpEF was detected in 10% of participants (n=76), and 10% were non-diagnostic. The participants with (vs. without) AI-HFpEF phenotype were older and had a higher burden of CVD risk factors, left ventricular hypertrophy, and left atrial dilation. Subclinical HFpEF, as defined by the gold standard criteria, was present in 5.3% of participants. The AUROC for diagnosis of subclinical HFpEF using the AI-HFpEF algorithm was 0.85 vs. 0.78 by the H2FpEF score (see Figure). In the adjusted analysis, the AI-HFpEF phenotype was significantly associated with greater odds of subclinical HFpEF by exercise and echocardiographic phenotyping criteria, higher E/e’ with exercise, lower VO2 peak, and worse LV and LA strain (see Table). In decision curve analysis, the AI-HFpEF algorithm identified 23 additional cases of subclinical HFpEF per 1000 screened participants compared with the H2FpEF score. Conclusion The AI-HFpEF algorithm can reliably identify community-dwelling individuals with subclinical HFpEF characterized by diastolic dysfunction at rest or exercise and impaired exercise capacity. Future studies are needed to assess the utility of the AI-HFpEF algorithm in screening for subclinical HFpEF in health systems.

Interethnic validation of an artificial intelligence algorithm for prediction of atrial fibrillation using sinus rhythm electrocardiogram

Abstract Background Previous research has demonstrated acceptable diagnostic accuracy of artificial intelligence (AI)-enabled sinus rhythm (SR) electrocardiogram (ECG) interpretation for identifying paroxysmal atrial fibrillation (AF). However, interethnic validations of the AI algorithms have not been widely conducted. Purpose We aimed to develop our own AI model for the identification of paroxysmal AF based on SR ECGs in the Korean population and to validate its diagnostic performance in Brazilian citizens. Methods We trained a Transformer-based vision network on 90% of a dataset comprising 808,194 ECGs from 121,282 patients at Seoul National University Bundang Hospital (2003-2020). The remaining 10% of the dataset was used for internal validation. The model was trained to compute a risk score for paroxysmal AF or new-onset AF within 2 years. External validation was conducted using non-AF ECGs from the CODE 15% dataset provided by the Telehealth Network of Minas Gerais, Brazil. Results In internal validation, our AI model achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.911 (95% CI: 0.902 - 0.921), with a sensitivity of 78.1% and a specificity of 89.0%. Subgroup analyses showed an AUROC of 0.893 (95% CI: 0.876 - 0.909) for patients in routine health checkups or outpatient settings, and 0.854 for patients with "Normal ECG" interpretations. In external validation with the CODE 15% dataset, the AI model exhibited an AUROC of 0.884 (95% CI: 0.869-0.900), which increased to 0.906 (95% CI: 0.893-0.919) when adjusted for age and sex (Figure 1). In the subset of patients with "Normal ECG" interpretations, the AUROC was 0.826 (95% CI: 0.769-0.883), increasing to 0.861 (95% CI: 0.814-0.908) after applying the same adjustments. Conclusions Our AI-powered SR ECG interpretation model demonstrated excellent diagnostic performance in predicting paroxysmal AF, with valid performance in the Brazilian population as well. This suggests that the model has potential for broad application across different ethnic groups.Figure 1

Detection of ATTR cardiac amyloidosis using a novel artificial intelligence algorithm for wearable-adapted noisy single-lead electrocardiograms

Abstract Background Amyloid cardiomyopathy (ATTR-CM) is often underdiagnosed, with very few diagnosed before the onset of symptoms, leading to delayed use of disease-modifying therapies that could have substantial prognostic implications early in the disease course. In the US, the disease has a substantially higher prevalence among Black adults, who are further challenged by lower access to healthcare services and lower preventative care, further reducing the ability for clinician-led diagnosis. We sought to develop an approach for identifying ATTR-CM on noisy single-lead ECGs obtainable on portable and wearable ECG devices, allowing broad use in community screening. Purpose To develop a portable/wearable device-adapted AI-ECG algorithm capable of detecting ATTR-CM on noisy single-lead ECGs. Methods We identified patients with ATTR-CM from 2015 through June 2023 across 5 hospitals of a large U.S.-based hospital system using positive bone scintigraphy scans or pharmacotherapy with an approved transthyretin stabilizer. In the development cohort, we used lead I of 12 lead ECGs, commonly captured by portable and wearable ECG devices, to train a novel noise-adapted model explicitly designed to infer information against simulated real-world noises. For this, ECG signals were augmented using random Gaussian real-world noise within four frequency ranges at multiple signal-to-noise ratios. We tested the model in an independent set of cases and matched controls drawn from July through December 2023. Results The development cohort consisted of 1,011 ECGs from 234 patients (median age 79 years [IQR:70-85], 17% women), with 10:1 age- and sex-matched 10,110 controls with 10,110 ECGs (age 79 [IQR:70-85] years 17.7% female). In the independent test set of 139 ECGs from 47 patients (age 80 [75-86] years, 32% women) and matched controls (1390 ECGs and patients) from July through December 2023, the AUROC (area under the receiver operating characteristic curve) of the AI-ECG model for ATTR-CM was 0.90 [95%CI: 0.88-0.92] (A). The AUPRC was 0.44 [0.36-0.53]. The sensitivity and specificity of the model were 0.85 and 0.80. The positive predictive value ranged from 0.12 at a 3% prevalence level to 0.59 at a 25% prevalence (B). Conclusions A novel portable and wearable-devices adapted model demonstrates promising performance for detecting ATTR-CM on single-lead ECGs, representing a more accessible method for screening in community-dwelling adults.

Diagnostic and prognostic evaluation of an echocardiography-based artificial intelligence algorithm for detecting HFpEF: a case-control analysis

Abstract Background Heart failure (HF) with preserved Ejection Fraction (HFpEF) is common among older adults and is associated with a high burden of morbidity and mortality. Despite its increasing prevalence, the diagnosis of HFpEF remains challenging. Recently, the FDA cleared an echocardiography-based AI HFpEF model that utilizes a 3-dimensional convolutional neural network to detect HFpEF using a single 4-chamber clip from a resting echocardiogram. However, the external validation of this algorithm against clinically adjudicated and confirmed HFpEF is limited. Purpose To evaluate the diagnostic and prognostic performance of the echocardiography-based AI HFpEF model in a cohort of HFpEF patients and matched controls. Methods The study included patients with clinically adjudicated HFpEF based on clinical history, normal ejection fraction (>45%), and evidence of elevated filling pressure by resting (PCWP > 15 mm Hg) or exercise invasive hemodynamics (PCWP > 25 mm Hg) or echocardiogram (E/e’ >14). The controls were age, sex, and BMI-matched patients without HF and a normal echocardiogram. The performance of the AI HFpEF model was evaluated using receiver operator curves. In patients with HFpEF, the association of the AI-HFpEF phenotype with elevated resting/exercise PCWP and peak exercise oxygen uptake (VO2peak) was assessed using multivariable logistic and linear regression models adjusting for age, sex, race, BMI, and comorbidities (diabetes, hypertension, kidney disease, atrial fibrillation). Results Of the 166 patients referred for evaluation of HFpEF, 82% had clinically adjudicated HFpEF, and 69.8% had elevated LV filling pressure at rest or exercise. In the matched cohorts of patients with clinically adjudicated HFpEF and matched control individuals (N = 122 each), the AI algorithm-based probability of HFpEF demonstrated good performance in identifying clinically adjudicated and hemodynamically confirmed HFpEF (AUROC: 0.75 for each) that was greater than the widely used H2FpEF score (0.69 and 0.70, Figure). In the HFpEF referral cohort, a higher probability of HFpEF based on the AI-algorithm phenotype was significantly associated with lower VO2peak (b [95% CI] per 5% higher probability: -0.11 [-0.21 to -0.01, P-value: 0.03] and greater odds of elevated PCWP (Odds ratio [95% CI] per 5% higher probability: 1.07 [1.01 – 1.15, P-value: 0.04] at rest or exercise after accounting for other confounders. Based on Youden’s index, the AI algorithm-based probability threshold of >0.75 was identified as the optimal cutoff for detecting HFpEF by the AI algorithm, with high sensitivity (0.85) and accuracy (0.74) and, adequate specificity (0.66). Conclusion The echocardiography-based AI HFpEF model demonstrated excellent sensitivity and discrimination in identifying patients with vs. without clinical HFpEF. Furthermore, the AI HFpEF model also had prognostic utility such that individuals with a higher probability of AI-HFpEF phenotype had more severe HFpEF.