Adv Protein Chem Struct Biol. 2024;139:263-288. doi: 10.1016/bs.apcsb.2023.11.015. Epub 2024 Feb 19.

ABSTRACT

In the past three decades, interest in using carbon-based nanomaterials (CBNs) in biomedical application has witnessed remarkable growth. Despite the rapid advancement, the translation of laboratory experimentation to clinical applications of nanomaterials is one of the major challenges. This might be attributed to poor understanding of bio-nano interface. Arguably, the most significant barrier is the complexity that arises by interplay of several factors like properties of nanomaterial (shape, size, surface chemistry), its interaction with suspending media (surface hydration and dehydration, surface reconstruction and release of free surface energy) and the interaction with biomolecules (conformational change in biomolecules, interaction with membrane and receptor). Tailoring a nanomaterial that minimally interacts with protein and lipids in the medium while effectively acts on target site in biological milieu has been very difficult. Computational methods and artificial intelligence techniques have displayed potential in effectively addressing this problem. Through predictive modelling and deep learning, computer-based methods have demonstrated the capability to create accurate models of interactions between nanoparticles and cell membranes, as well as the uptake of nanomaterials by cells. Computer-based simulations techniques enable these computational models to forecast how making particular alterations to a material's physical and chemical properties could enhance functional aspects, such as the retention of drugs, the process of cellular uptake and biocompatibility. We review the most recent progress regarding the bio-nano interface studies between the plasma proteins and CBNs with a special focus on computational simulations based on molecular dynamics and density functional theory.

PMID:38448137 | DOI:10.1016/bs.apcsb.2023.11.015

Adv Protein Chem Struct Biol. 2024;139:141-171. doi: 10.1016/bs.apcsb.2023.11.007. Epub 2023 Dec 20.

ABSTRACT

Advancements in genome sequencing have expanded the scope of investigating mutations in proteins across different diseases. Amino acid mutations in a protein alter its structure, stability and function and some of them lead to diseases. Identification of disease-causing mutations is a challenging task and it will be helpful for designing therapeutic strategies. Hence, mutation data available in the literature have been curated and stored in several databases, which have been effectively utilized for developing computational methods to identify deleterious mutations (drivers), using sequence and structure-based properties of proteins. In this chapter, we describe the contents of specific databases that have information on disease-causing and neutral mutations followed by sequence and structure-based properties. Further, characteristic features of disease-causing mutations will be discussed along with computational methods for identifying cancer hotspot residues and disease-causing mutations in proteins.

PMID:38448134 | DOI:10.1016/bs.apcsb.2023.11.007

Neuroimage. 2024 Mar 4:120559. doi: 10.1016/j.neuroimage.2024.120559. Online ahead of print.

ABSTRACT

As the field of computational cognitive neuroscience continues to expand and generate new theories, there is a growing need for more advanced methods to test the hypothesis of brain-behavior relationships. Recent progress in Bayesian cognitive modeling has enabled the combination of neural and behavioral models into a single unifying framework. However, these approaches require manual feature extraction, and lack the capability to discover previously unknown neural features in more complex data. Consequently, this would hinder the expressiveness of the models. To address these challenges, we propose a Neurocognitive Variational Autoencoder (NCVA) to conjoin high-dimensional EEG with a cognitive model in both generative and predictive modeling analyses. Importantly, our NCVA enables both the prediction of EEG signals given behavioral data and the estimation of cognitive model parameters from EEG signals. This novel approach can allow for a more comprehensive understanding of the triplet relationship between behavior, brain activity, and cognitive processes.

PMID:38447682 | DOI:10.1016/j.neuroimage.2024.120559

Am J Obstet Gynecol MFM. 2024 Mar 4:101337. doi: 10.1016/j.ajogmf.2024.101337. Online ahead of print.

ABSTRACT

BACKGROUND: This study used electrocardiogram (ECG) data in conjunction with artificial intelligence (AI) methods as a non-invasive tool for detecting peripartum cardiomyopathy (PPCM).

OBJECTIVE: The primary objective was to assess the efficacy of a heart failure detection model for detecting peripartum cardiomyopathy detection using an AI deep learning model called a 1-dimensional convolutional neural network.

STUDY DESIGN: We first built a deep learning model for heart failure detection using retrospective data at University of Tennessee Health Science Center (UTHSC). Cases were adult and non-pregnant females with a heart failure diagnosis; controls were adult non-pregnant females without heart failure. The model was then tested on an independent cohort of pregnant women at UTHSC who either did or did not have peripartum cardiomyopathy. We also tested the model in an external cohort of pregnant women at Atrium Health Wake Forest Baptist (AHWFB). Key outcomes were assessed using the area under the receiver operating characteristic curve (AUC). We also repeated our analysis using only lead I ECG as an input to assess feasibility of remote monitoring via wearables that can capture single-lead ECG data.

RESULTS: The UTHSC heart failure cohort comprised 346,339 ECGs from 142,601 patients. In this cohort, 60% were Black and 37% were white, with an average age (SD) of 53 (19). The heart failure detection model achieved an AUC of 0.92 on the holdout set. We then tested the ability of the heart failure model to detect peripartum cardiomyopathy in an independent cohort of pregnant women from UTHSC and an external cohort of pregnant women from AWFBH. The independent UTHSC cohort included 158 ECGs from 115 patients; our deep learning model achieved an AUC of 0.83 [0.77-0.89] for this dataset. The external AHWFB cohort involved 80 ECGs from 43 patients; our deep learning model achieved an AUC of 0.83 [0.77-0.89] AUC of 0.94 [0.91-0.98] for this dataset. For identifying peripartum cardiomyopathy diagnosed 10 or more days post-delivery, the model achieved an AUC of 0.88 [0.81-0.94] for the UTHSC cohort and an AUC of 0.96 [0.93-0.99] for the AHWFB cohort. When we repeated our analysis by building a heart failure detection model using only lead I ECGs, we obtained similarly high detection accuracies, with AUCs of 0.73 and 0.93 for the UTHSC and AHWFB cohorts, respectively.

CONCLUSIONS: AI can accurately detect peripartum cardiomyopathy from ECG alone. A simple ECG-AI-based peripartum screening could result in a more timely diagnosis. Since results with 1-lead ECG data were similar to those obtained using all 12 leads, future studies will focus on remote screening for peripartum cardiomyopathy using smartwatches that can capture single-lead ECG data.

PMID:38447673 | DOI:10.1016/j.ajogmf.2024.101337

J Biomed Inform. 2024 Mar 4:104621. doi: 10.1016/j.jbi.2024.104621. Online ahead of print.

ABSTRACT

OBJECTIVE: The primary objective of this review is to investigate the effectiveness of machine learning and deep learning methodologies in the context of extracting adverse drug events (ADEs) from clinical benchmark datasets. We conduct an in-depth analysis, aiming to compare the merits and drawbacks of both machine learning and deep learning techniques, particularly within the framework of named-entity recognition (NER) and relation classification (RC) tasks related to ADE extraction. Additionally, our focus extends to the examination of specific features and their impact on the overall performance of these methodologies. In a broader perspective, our research extends to ADE extraction from various sources, including biomedical literature, social media data, and drug labels, removing the limitation to exclusively machine learning or deep learning methods.

METHODS: We conducted an extensive literature review on PubMed using the query "(((machine learning [Medical Subject Headings (MeSH) Terms]) OR (deep learning [MeSH Terms])) AND (adverse drug event [MeSH Terms])) AND (extraction)", and supplemented this with a snowballing approach to review 275 references sourced from retrieved articles.

RESULTS: In our analysis, we included twelve articles for review. For the NER task, deep learning models outperformed machine learning models. In the RC task, gradient Boosting, multilayer perceptron and random forest models excelled. The Bidirectional Encoder Representations from Transformers (BERT) model consistently achieved the best performance in the end-to-end task. Future efforts in the end-to-end task should prioritize improving NER accuracy, especially for 'ADE' and 'Reason'.

CONCLUSION: These findings hold significant implications for advancing the field of ADE extraction and pharmacovigilance, ultimately contributing to improved drug safety monitoring and healthcare outcomes.

PMID:38447600 | DOI:10.1016/j.jbi.2024.104621

Structure. 2024 Mar 4:S0969-2126(24)00043-1. doi: 10.1016/j.str.2024.02.004. Online ahead of print.

ABSTRACT

Identifying binding compounds against a target protein is crucial for large-scale virtual screening in drug development. Recently, network-based methods have been developed for compound-protein interaction (CPI) prediction. However, they are difficult to be applied to unseen (i.e., never-seen-before) proteins and compounds. In this study, we propose SgCPI to incorporate local known interacting networks to predict CPI interactions. SgCPI randomly samples the local CPI network of the query compound-protein pair as a subgraph and applies a heterogeneous graph neural network (HGNN) to embed the active/inactive message of the subgraph. For unseen compounds and proteins, SgCPI-KD takes SgCPI as the teacher model to distillate its knowledge by estimating the potential neighbors. Experimental results indicate: (1) the sampled subgraphs of the CPI network introduce efficient knowledge for unseen molecular prediction with the HGNNs, and (2) the knowledge distillation strategy is beneficial to the double-blind interaction prediction by estimating molecular neighbors and distilling knowledge.

PMID:38447575 | DOI:10.1016/j.str.2024.02.004

Comput Biol Med. 2024 Feb 29;171:108229. doi: 10.1016/j.compbiomed.2024.108229. Online ahead of print.

ABSTRACT

Conventional COVID-19 testing methods have some flaws: they are expensive and time-consuming. Chest X-ray (CXR) diagnostic approaches can alleviate these flaws to some extent. However, there is no accurate and practical automatic diagnostic framework with good interpretability. The application of artificial intelligence (AI) technology to medical radiography can help to accurately detect the disease, reduce the burden on healthcare organizations, and provide good interpretability. Therefore, this study proposes a new deep neural network (CNN) based on CXR for COVID-19 diagnosis - CodeNet. This method uses contrastive learning to make full use of latent image data to enhance the model's ability to extract features and generalize across different data domains. On the evaluation dataset, the proposed method achieves an accuracy as high as 94.20%, outperforming several other existing methods used for comparison. Ablation studies validate the efficacy of the proposed method, while interpretability analysis shows that the method can effectively guide clinical professionals. This work demonstrates the superior detection performance of a CNN using contrastive learning techniques on CXR images, paving the way for computer vision and artificial intelligence technologies to leverage massive medical data for disease diagnosis.

PMID:38447500 | DOI:10.1016/j.compbiomed.2024.108229

Comput Biol Med. 2024 Feb 22;171:108131. doi: 10.1016/j.compbiomed.2024.108131. Online ahead of print.

ABSTRACT

Morphological features of individual nuclei serve as a dependable foundation for pathologists in making accurate diagnoses. Existing methods that rely on spatial information for feature extraction have achieved commendable results in nuclei segmentation tasks. However, these approaches are not sufficient to extract edge information of nuclei with small sizes and blurred outlines. Moreover, the lack of attention to the interior of the nuclei leads to significant internal inconsistencies. To address these challenges, we introduce a novel Spatial-Frequency Enhancement Network (SFE-Net) to incorporate spatial-frequency features and promote intra-nuclei consistency for robust nuclei segmentation. Specifically, SFE-Net incorporates a distinctive Spatial-Frequency Feature Extraction module and a Spatial-Guided Feature Enhancement module, which are designed to preserve spatial-frequency information and enhance feature representation respectively, to achieve comprehensive extraction of edge information. Furthermore, we introduce the Label-Guided Distillation method, which utilizes semantic features to guide the segmentation network in strengthening boundary constraints and learning the intra-nuclei consistency of individual nuclei, to improve the robustness of nuclei segmentation. Extensive experiments on three publicly available histopathology image datasets (MoNuSeg, TNBC and CryoNuSeg) demonstrate the superiority of our proposed method, which achieves 79.23%, 81.96% and 73.26% Aggregated Jaccard Index, respectively. The proposed model is available at https://github.com/jinshachen/SFE-Net.

PMID:38447498 | DOI:10.1016/j.compbiomed.2024.108131

Eur J Radiol. 2024 Mar 2;174:111405. doi: 10.1016/j.ejrad.2024.111405. Online ahead of print.

ABSTRACT

PURPOSE: Clinical risk scores are essential for predicting outcomes in stroke patients. The advancements in deep learning (DL) techniques provide opportunities to develop prediction applications using magnetic resonance (MR) images. We aimed to develop an MR-based DL imaging biomarker for predicting outcomes in acute ischemic stroke (AIS) and evaluate its additional benefit to current risk scores.

METHOD: This study included 3338 AIS patients. We trained a DL model using deep neural network architectures on MR images and radiomics to predict poor functional outcomes at three months post-stroke. The DL model generated a DL score, which served as the DL imaging biomarker. We compared the predictive performance of this biomarker to five risk scores on a holdout test set. Additionally, we assessed whether incorporating the imaging biomarker into the risk scores improved the predictive performance.

RESULTS: The DL imaging biomarker achieved an area under the receiver operating characteristic curve (AUC) of 0.788. The AUCs of the five studied risk scores were 0.789, 0.793, 0.804, 0.810, and 0.826, respectively. The imaging biomarker's predictive performance was comparable to four of the risk scores but inferior to one (p = 0.038). Adding the imaging biomarker to the risk scores improved the AUCs (p-values) to 0.831 (0.003), 0.825 (0.001), 0.834 (0.003), 0.836 (0.003), and 0.839 (0.177), respectively. The net reclassification improvement and integrated discrimination improvement indices also showed significant improvements (all p < 0.001).

CONCLUSIONS: Using DL techniques to create an MR-based imaging biomarker is feasible and enhances the predictive ability of current risk scores.

PMID:38447430 | DOI:10.1016/j.ejrad.2024.111405

ACS Appl Mater Interfaces. 2024 Mar 6. doi: 10.1021/acsami.3c15956. Online ahead of print.

ABSTRACT

Driver assistance systems can help drivers achieve better control of their vehicles while driving and reduce driver fatigue and errors. However, the current driver assistance devices have a complex structure and severely violate the privacy of drivers, hindering the development of driver assistance technology. To address these limitations, this article proposes an intelligent driver assistance monitoring system (IDAMS), which combines a Kresling origami structure-based triboelectric sensor (KOS-TS) and a convolutional neural network (CNN)-based data analysis. For different driving behaviors, the output signals of the KOS-TSs contain various features, such as a driver's pressing force, pressing time, and sensor triggering sequence. This study develops a multiscale CNN that employs different pooling methods to process KOS-TS data and analyze temporal information. The proposed IDAMS is verified by driver identification experiments, and the results show that the accuracy of the IDAMS in discriminating eight different users is improved from 96.25% to 99.38%. In addition, the results indicate that IDAMS can successfully monitor driving behaviors and can accurately distinguish between different driving behaviors. Finally, the proposed IDAMS has excellent hands-off detection (HOD), identification, and driving behavior monitoring capabilities and shows broad potential for application in the fields of safety warning, personalization, and human-computer interaction.

PMID:38447140 | DOI:10.1021/acsami.3c15956

Sci Adv. 2024 Mar 8;10(10):eadk6669. doi: 10.1126/sciadv.adk6669. Epub 2024 Mar 6.

ABSTRACT

Environmental hazard assessments are reliant on toxicity data that cover multiple organism groups. Generating experimental toxicity data is, however, resource-intensive and time-consuming. Computational methods are fast and cost-efficient alternatives, but the low accuracy and narrow applicability domains have made their adaptation slow. Here, we present a AI-based model for predicting chemical toxicity. The model uses transformers to capture toxicity-specific features directly from the chemical structures and deep neural networks to predict effect concentrations. The model showed high predictive performance for all tested organism groups-algae, aquatic invertebrates and fish-and has, in comparison to commonly used QSAR methods, a larger applicability domain and a considerably lower error. When the model was trained on data with multiple effect concentrations (EC50/EC10), the performance was further improved. We conclude that deep learning and transformers have the potential to markedly advance computational prediction of chemical toxicity.

PMID:38446886 | DOI:10.1126/sciadv.adk6669

Sci Adv. 2024 Mar 8;10(10):eadk2298. doi: 10.1126/sciadv.adk2298. Epub 2024 Mar 6.

ABSTRACT

Adoptive T cell therapies rely on the production of T cells with an antigen receptor that directs their specificity toward tumor-specific antigens. Methods for identifying relevant T cell receptor (TCR) sequences, predominantly achieved through the enrichment of antigen-specific T cells, represent a major bottleneck in the production of TCR-engineered cell therapies. Fluctuation of intracellular calcium is a proximal readout of TCR signaling and candidate marker for antigen-specific T cell identification that does not require T cell expansion; however, calcium fluctuations downstream of TCR engagement are highly variable. We propose that machine learning algorithms may allow for T cell classification from complex datasets such as polyclonal T cell signaling events. Using deep learning tools, we demonstrate accurate prediction of TCR-transgenic CD8+ T cell activation based on calcium fluctuations and test the algorithm against T cells bearing a distinct TCR as well as polyclonal T cells. This provides the foundation for an antigen-specific TCR sequence identification pipeline for adoptive T cell therapies.

PMID:38446885 | DOI:10.1126/sciadv.adk2298

Sci Adv. 2024 Mar 8;10(10):eadk3854. doi: 10.1126/sciadv.adk3854. Epub 2024 Mar 6.

ABSTRACT

Liquid crystal elastomers (LCEs) have garnered attention for their remarkable reversible strains under various stimuli. Early studies on LCEs mainly focused on basic dimensional changes in macrostructures or quasi-three-dimensional (3D) microstructures. However, fabricating complex 3D microstructures and cross-scale LCE-based structures has remained challenging. In this study, we report a compatible method named melt electrowriting (MEW) to fabricate LCE-based microfiber actuators and various 3D actuators on the micrometer to centimeter scales. By controlling printing parameters, these actuators were fabricated with high resolutions (4.5 to 60 μm), actuation strains (10 to 55%), and a maximum work density of 160 J/kg. In addition, through the integration of a deep learning-based model, we demonstrated the application of LCE materials in temperature field sensing. Large-scale, real-time, LCE grid-based spatial temperature field sensors have been designed, exhibiting a low response time of less than 42 ms and a high precision of 94.79%.

PMID:38446880 | DOI:10.1126/sciadv.adk3854

Int J Surg. 2024 Mar 4. doi: 10.1097/JS9.0000000000001274. Online ahead of print.

NO ABSTRACT

PMID:38446840 | DOI:10.1097/JS9.0000000000001274

PLoS One. 2024 Mar 6;19(3):e0299265. doi: 10.1371/journal.pone.0299265. eCollection 2024.

ABSTRACT

Computer-aided diagnosis systems based on deep learning algorithms have shown potential applications in rapid diagnosis of diabetic retinopathy (DR). Due to the superior performance of Transformer over convolutional neural networks (CNN) on natural images, we attempted to develop a new model to classify referable DR based on a limited number of large-size retinal images by using Transformer. Vision Transformer (ViT) with Masked Autoencoders (MAE) was applied in this study to improve the classification performance of referable DR. We collected over 100,000 publicly fundus retinal images larger than 224×224, and then pre-trained ViT on these retinal images using MAE. The pre-trained ViT was applied to classify referable DR, the performance was also compared with that of ViT pre-trained using ImageNet. The improvement in model classification performance by pre-training with over 100,000 retinal images using MAE is superior to that pre-trained with ImageNet. The accuracy, area under curve (AUC), highest sensitivity and highest specificity of the present model are 93.42%, 0.9853, 0.973 and 0.9539, respectively. This study shows that MAE can provide more flexibility to the input image and substantially reduce the number of images required. Meanwhile, the pretraining dataset scale in this study is much smaller than ImageNet, and the pre-trained weights from ImageNet are not required also.

PMID:38446810 | DOI:10.1371/journal.pone.0299265

Brief Bioinform. 2024 Jan 22;25(2):bbae076. doi: 10.1093/bib/bbae076.

ABSTRACT

Identifying protein-protein interactions (PPIs) is crucial for deciphering biological pathways. Numerous prediction methods have been developed as cheap alternatives to biological experiments, reporting surprisingly high accuracy estimates. We systematically investigated how much reproducible deep learning models depend on data leakage, sequence similarities and node degree information, and compared them with basic machine learning models. We found that overlaps between training and test sets resulting from random splitting lead to strongly overestimated performances. In this setting, models learn solely from sequence similarities and node degrees. When data leakage is avoided by minimizing sequence similarities between training and test set, performances become random. Moreover, baseline models directly leveraging sequence similarity and network topology show good performances at a fraction of the computational cost. Thus, we advocate that any improvements should be reported relative to baseline methods in the future. Our findings suggest that predicting PPIs remains an unsolved task for proteins showing little sequence similarity to previously studied proteins, highlighting that further experimental research into the 'dark' protein interactome and better computational methods are needed.

PMID:38446741 | DOI:10.1093/bib/bbae076

Brief Bioinform. 2024 Jan 22;25(2):bbae078. doi: 10.1093/bib/bbae078.

ABSTRACT

Antimicrobial peptides (AMPs), short peptides with diverse functions, effectively target and combat various organisms. The widespread misuse of chemical antibiotics has led to increasing microbial resistance. Due to their low drug resistance and toxicity, AMPs are considered promising substitutes for traditional antibiotics. While existing deep learning technology enhances AMP generation, it also presents certain challenges. Firstly, AMP generation overlooks the complex interdependencies among amino acids. Secondly, current models fail to integrate crucial tasks like screening, attribute prediction and iterative optimization. Consequently, we develop a integrated deep learning framework, Diff-AMP, that automates AMP generation, identification, attribute prediction and iterative optimization. We innovatively integrate kinetic diffusion and attention mechanisms into the reinforcement learning framework for efficient AMP generation. Additionally, our prediction module incorporates pre-training and transfer learning strategies for precise AMP identification and screening. We employ a convolutional neural network for multi-attribute prediction and a reinforcement learning-based iterative optimization strategy to produce diverse AMPs. This framework automates molecule generation, screening, attribute prediction and optimization, thereby advancing AMP research. We have also deployed Diff-AMP on a web server, with code, data and server details available in the Data Availability section.

PMID:38446739 | DOI:10.1093/bib/bbae078

Brief Bioinform. 2024 Jan 22;25(2):bbae081. doi: 10.1093/bib/bbae081.

ABSTRACT

Accurately predicting the binding affinity between proteins and ligands is crucial in drug screening and optimization, but it is still a challenge in computer-aided drug design. The recent success of AlphaFold2 in predicting protein structures has brought new hope for deep learning (DL) models to accurately predict protein-ligand binding affinity. However, the current DL models still face limitations due to the low-quality database, inaccurate input representation and inappropriate model architecture. In this work, we review the computational methods, specifically DL-based models, used to predict protein-ligand binding affinity. We start with a brief introduction to protein-ligand binding affinity and the traditional computational methods used to calculate them. We then introduce the basic principles of DL models for predicting protein-ligand binding affinity. Next, we review the commonly used databases, input representations and DL models in this field. Finally, we discuss the potential challenges and future work in accurately predicting protein-ligand binding affinity via DL models.

PMID:38446737 | DOI:10.1093/bib/bbae081

IEEE J Biomed Health Inform. 2024 Mar;28(3):1528-1539. doi: 10.1109/JBHI.2024.3350247.

ABSTRACT

Colorectal cancer is a prevalent and life-threatening disease, where colorectal cancer liver metastasis (CRLM) exhibits the highest mortality rate. Currently, surgery stands as the most effective curative option for eligible patients. However, due to the insufficient performance of traditional methods and the lack of multi-modality MRI feature complementarity in existing deep learning methods, the prognosis of CRLM surgical resection has not been fully explored. This paper proposes a new method, multi-modal guided complementary network (MGCNet), which employs multi-sequence MRI to predict 1-year recurrence and recurrence-free survival in patients after CRLM resection. In light of the complexity and redundancy of features in the liver region, we designed the multi-modal guided local feature fusion module to utilize the tumor features to guide the dynamic fusion of prognostically relevant local features within the liver. On the other hand, to solve the loss of spatial information during multi-sequence MRI fusion, the cross-modal complementary external attention module designed an external mask branch to establish inter-layer correlation. The results show that the model has accuracy (ACC) of 0.79, the area under the curve (AUC) of 0.84, C-Index of 0.73, and hazard ratio (HR) of 4.0, which is a significant improvement over state-of-the-art methods. Additionally, MGCNet exhibits good interpretability.

PMID:38446655 | DOI:10.1109/JBHI.2024.3350247

JAMA Cardiol. 2024 Mar 6. doi: 10.1001/jamacardio.2024.0039. Online ahead of print.

ABSTRACT

IMPORTANCE: Congenital long QT syndrome (LQTS) is associated with syncope, ventricular arrhythmias, and sudden death. Half of patients with LQTS have a normal or borderline-normal QT interval despite LQTS often being detected by QT prolongation on resting electrocardiography (ECG).

OBJECTIVE: To develop a deep learning-based neural network for identification of LQTS and differentiation of genotypes (LQTS1 and LQTS2) using 12-lead ECG.

DESIGN, SETTING, AND PARTICIPANTS: This diagnostic accuracy study used ECGs from patients with suspected inherited arrhythmia enrolled in the Hearts in Rhythm Organization Registry (HiRO) from August 2012 to December 2021. The internal dataset was derived at 2 sites and an external validation dataset at 4 sites within the HiRO Registry; an additional cross-sectional validation dataset was from the Montreal Heart Institute. The cohort with LQTS included probands and relatives with pathogenic or likely pathogenic variants in KCNQ1 or KCNH2 genes with normal or prolonged corrected QT (QTc) intervals.

EXPOSURES: Convolutional neural network (CNN) discrimination between LQTS1, LQTS2, and negative genetic test results.

MAIN OUTCOMES AND MEASURES: The main outcomes were area under the curve (AUC), F1 scores, and sensitivity for detecting LQTS and differentiating genotypes using a CNN method compared with QTc-based detection.

RESULTS: A total of 4521 ECGs from 990 patients (mean [SD] age, 42 [18] years; 589 [59.5%] female) were analyzed. External validation within the national registry (101 patients) demonstrated the CNN's high diagnostic capacity for LQTS detection (AUC, 0.93; 95% CI, 0.89-0.96) and genotype differentiation (AUC, 0.91; 95% CI, 0.86-0.96). This surpassed expert-measured QTc intervals in detecting LQTS (F1 score, 0.84 [95% CI, 0.78-0.90] vs 0.22 [95% CI, 0.13-0.31]; sensitivity, 0.90 [95% CI, 0.86-0.94] vs 0.36 [95% CI, 0.23-0.47]), including in patients with normal or borderline QTc intervals (F1 score, 0.70 [95% CI, 0.40-1.00]; sensitivity, 0.78 [95% CI, 0.53-0.95]). In further validation in a cross-sectional cohort (406 patients) of high-risk patients and genotype-negative controls, the CNN detected LQTS with an AUC of 0.81 (95% CI, 0.80-0.85), which was better than QTc interval-based detection (AUC, 0.74; 95% CI, 0.69-0.78).

CONCLUSIONS AND RELEVANCE: The deep learning model improved detection of congenital LQTS from resting ECGs and allowed for differentiation between the 2 most common genetic subtypes. Broader validation over an unselected general population may support application of this model to patients with suspected LQTS.

PMID:38446445 | DOI:10.1001/jamacardio.2024.0039