Deep learning-based long-term risk evaluation of incident type 2 diabetes using electrocardiogram in a non-diabetic population: a retrospective, multicentre study.

Background: Diabetes is a major public health concern. We aimed to evaluate the long-term risk of incident type 2 diabetes in a non-diabetic population using a deep learning model (DLM) detecting prevalent type 2 diabetes using electrocardiogram (ECG). Methods: In this retrospective study, participants who underwent health checkups at two tertiary hospitals in Seoul, South Korea, between Jan 1, 2001 and Dec 31, 2022 were included. Type 2 diabetes was defined as glucose ≥126 mg/dL or glycated haemoglobin (HbA1c) ≥ 6.5%. For survival analysis on incident type 2 diabetes, we introduced an additional variable, diabetic ECG, which is determined by the DLM trained on ECG and corresponding prevalent diabetes. It was assumed that non-diabetic individuals with diabetic ECG had a higher risk of incident type 2 diabetes than those with non-diabetic ECG. The one-dimensional ResNet-based model was adopted for the DLM, and the Guided Grad-CAM was used to localise important regions of ECG. We divided the non-diabetic group into the diabetic ECG group (false positive) and the non-diabetic ECG (true negative) group according to the DLM decision, and performed a Cox proportional hazard model, considering the occurrence of type 2 diabetes more than six months after the visit. Findings: 190,581 individuals were included in the study with a median follow-up period of 11.84 years. The areas under the receiver operating characteristic curve for prevalent type 2 diabetes detection were 0.816 (0.807-0.825) and 0.762 (0.754-0.770) for the internal and external validations, respectively. The model primarily focused on the QRS duration and, occasionally, P or T waves. The diabetic ECG group exhibited an increased risk of incident type 2 diabetes compared with the non-diabetic ECG group, with hazard ratios of 2.15 (1.82-2.53) and 1.92 (1.74-2.11) for internal and external validation, respectively. Interpretation: In the non-diabetic group, those whose ECG was classified as diabetes by the DLM were at a higher risk of incident type 2 diabetes than those whose ECG was not. Additional clinical research on the relationship between the phenotype of ECG and diabetes to support the results and further investigation with tracked data and various ECG recording systems are suggested for future works. Funding: National Research Foundation of Korea.

Real-time machine learning model to predict in-hospital cardiac arrest using heart rate variability in ICU.

Predicting in-hospital cardiac arrest in patients admitted to an intensive care unit (ICU) allows prompt interventions to improve patient outcomes. We developed and validated a machine learning-based real-time model for in-hospital cardiac arrest predictions using electrocardiogram (ECG)-based heart rate variability (HRV) measures. The HRV measures, including time/frequency domains and nonlinear measures, were calculated from 5 min epochs of ECG signals from ICU patients. A light gradient boosting machine (LGBM) algorithm was used to develop the proposed model for predicting in-hospital cardiac arrest within 0.5-24 h. The LGBM model using 33 HRV measures achieved an area under the receiver operating characteristic curve of 0.881 (95% CI: 0.875-0.887) and an area under the precision-recall curve of 0.104 (95% CI: 0.093-0.116). The most important feature was the baseline width of the triangular interpolation of the RR interval histogram. As our model uses only ECG data, it can be easily applied in clinical practice.

Automatic segmentation of atrial fibrillation and flutter in single-lead electrocardiograms by self-supervised learning and Transformer architecture.

OBJECTIVES: Automatic detection of atrial fibrillation and flutter (AF/AFL) is a significant concern in preventing stroke and mitigating hemodynamic instability. Herein, we developed a Transformer-based deep learning model for AF/AFL segmentation in single-lead electrocardiograms (ECGs) by self-supervised learning with masked signal modeling (MSM). MATERIALS AND METHODS: We retrieved data from 11 open-source databases on PhysioNet; 7 of these databases included labeled ECGs, while the other 4 were without labels. Each database contained ECG recordings with durations of ≥30 s. A total of 24 intradialytic ECGs with paroxysmal AF/AFL during 4 h of hemodialysis sessions at Seoul National University Hospital were used for external validation. The model was pretrained by predicting masked areas of ECG signals and fine-tuned by predicting AF/AFL areas. Cross-database validation was used for evaluation, and the intersection over union (IOU) was used as a main performance metric in external database validation. RESULTS: In the 7 labeled databases, the areas marked as AF/AFL constituted 41.1% of the total ECG signals, ranging from 0.19% to 51.31%. In the evaluation per ECG segment, the model achieved IOU values of 0.9254 and 0.9477 for AF/AFL segmentation and other segmentation tasks, respectively. When applied to intradialytic ECGs with paroxysmal AF/AFL, the IOUs for the segmentation of AF/AFL and non-AF/AFL were 0.9896 and 0.9650, respectively. Model performance by different training procedure indicated that pretraining with MSM and the application of an appropriate masking ratio both contributed to the model performance. It also showed higher IOUs of AF/AFL labels than in previous studies when training and test databases were matched. CONCLUSION: The present model with self-supervised learning by MSM performs robustly in segmenting AF/AFL.

Real-time dual prediction of intradialytic hypotension and hypertension using an explainable deep learning model.

Both intradialytic hypotension (IDH) and hypertension (IDHTN) are associated with poor outcomes in hemodialysis patients, but a model predicting dual outcomes in real-time has never been developed. Herein, we developed an explainable deep learning model with a sequence-to-sequence-based attention network to predict both of these events simultaneously. We retrieved 302,774 hemodialysis sessions from the electronic health records of 11,110 patients, and these sessions were split into training (70%), validation (10%), and test (20%) datasets through patient randomization. The outcomes were defined when nadir systolic blood pressure (BP) < 90 mmHg (termed IDH-1), a decrease in systolic BP ≥ 20 mmHg and/or a decrease in mean arterial pressure ≥ 10 mmHg (termed IDH-2), or an increase in systolic BP ≥ 10 mmHg (i.e., IDHTN) occurred within 1 h. We developed a temporal fusion transformer (TFT)-based model and compared its performance in the test dataset, including receiver operating characteristic curve (AUROC) and area under the precision-recall curves (AUPRC), with those of other machine learning models, such as recurrent neural network, light gradient boosting machine, random forest, and logistic regression. Among all models, the TFT-based model achieved the highest AUROCs of 0.953 (0.952-0.954), 0.892 (0.891-0.893), and 0.889 (0.888-0.890) in predicting IDH-1, IDH-2, and IDHTN, respectively. The AUPRCs in the TFT-based model for these outcomes were higher than the other models. The factors that contributed the most to the prediction were age and previous session, which were time-invariant variables, as well as systolic BP and elapsed time, which were time-varying variables. The present TFT-based model predicts both IDH and IDHTN in real time and offers explainable variable importance.

Machine learning-based prediction of intraoperative hypoxemia for pediatric patients.

BACKGROUND: Reducing the duration of intraoperative hypoxemia in pediatric patients by means of rapid detection and early intervention is considered crucial by clinicians. We aimed to develop and validate a machine learning model that can predict intraoperative hypoxemia events 1 min ahead in children undergoing general anesthesia. METHODS: This retrospective study used prospectively collected intraoperative vital signs and parameters from the anesthesia ventilator machine extracted every 2 s in pediatric patients undergoing surgery under general anesthesia between January 2019 and October 2020 in a tertiary academic hospital. Intraoperative hypoxemia was defined as oxygen saturation <95% at any point during surgery. Three common machine learning techniques were employed to develop models using the training dataset: gradient-boosting machine (GBM), long short-term memory (LSTM), and transformer. The performances of the models were compared using the area under the receiver operating characteristics curve using randomly assigned internal testing dataset. We also validated the developed models using temporal holdout dataset. Pediatric patient surgery cases between November 2020 and January 2021 were used. The performances of the models were compared using the area under the receiver operating characteristic curve (AUROC). RESULTS: In total, 1,540 (11.73%) patients with intraoperative hypoxemia out of 13,130 patients' records with 2,367 episodes were included for developing the model dataset. After model development, 200 (13.25%) of the 1,510 patients' records with 289 episodes were used for holdout validation. Among the models developed, the GBM had the highest AUROC of 0.904 (95% confidence interval [CI] 0.902 to 0.906), which was significantly higher than that of the LSTM (0.843, 95% CI 0.840 to 0.846 P < .001) and the transformer model (0.885, 95% CI, 0.882-0.887, P < .001). In holdout validation, GBM also demonstrated best performance with an AUROC of 0.939 (95% CI 0.936 to 0.941) which was better than LSTM (0.904, 95% CI 0.900 to 0.907, P < .001) and the transformer model (0.929, 95% CI 0.926 to 0.932, P < .001). CONCLUSIONS: Machine learning models can be used to predict upcoming intraoperative hypoxemia in real-time based on the biosignals acquired by patient monitors, which can be useful for clinicians for prediction and proactive treatment of hypoxemia in an intraoperative setting.

Artificial intelligence in perioperative medicine: a narrative review.

Recent advancements in artificial intelligence (AI) techniques have enabled the development of accurate prediction models using clinical big data. AI models for perioperative risk stratification, intraoperative event prediction, biosignal analyses, and intensive care medicine have been developed in the field of perioperative medicine. Some of these models have been validated using external datasets and randomized controlled trials. Once these models are implemented in electronic health record systems or software medical devices, they could help anesthesiologists improve clinical outcomes by accurately predicting complications and suggesting optimal treatment strategies in real-time. This review provides an overview of the AI techniques used in perioperative medicine and a summary of the studies that have been published using these techniques. Understanding these techniques will aid in their appropriate application in clinical practice.

Development and Validation of an Arterial Pressure-Based Cardiac Output Algorithm Using a Convolutional Neural Network: Retrospective Study Based on Prospective Registry Data.

BACKGROUND: Arterial pressure-based cardiac output (APCO) is a less invasive method for estimating cardiac output without concerns about complications from the pulmonary artery catheter (PAC). However, inaccuracies of currently available APCO devices have been reported. Improvements to the algorithm by researchers are impossible, as only a subset of the algorithm has been released. OBJECTIVE: In this study, an open-source algorithm was developed and validated using a convolutional neural network and a transfer learning technique. METHODS: A retrospective study was performed using data from a prospective cohort registry of intraoperative bio-signal data from a university hospital. The convolutional neural network model was trained using the arterial pressure waveform as input and the stroke volume (SV) value as the output. The model parameters were pretrained using the SV values from a commercial APCO device (Vigileo or EV1000 with the FloTrac algorithm) and adjusted with a transfer learning technique using SV values from the PAC. The performance of the model was evaluated using absolute error for the PAC on the testing dataset from separate periods. Finally, we compared the performance of the deep learning model and the FloTrac with the SV values from the PAC. RESULTS: A total of 2057 surgical cases (1958 training and 99 testing cases) were used in the registry. In the deep learning model, the absolute errors of SV were 14.5 (SD 13.4) mL (10.2 [SD 8.4] mL in cardiac surgery and 17.4 [SD 15.3] mL in liver transplantation). Compared with FloTrac, the absolute errors of the deep learning model were significantly smaller (16.5 [SD 15.4] and 18.3 [SD 15.1], P<.001). CONCLUSIONS: The deep learning-based APCO algorithm showed better performance than the commercial APCO device. Further improvement of the algorithm developed in this study may be helpful for estimating cardiac output accurately in clinical practice and optimizing high-risk patient care.

Automatic prediction of left cardiac chamber enlargement from chest radiographs using convolutional neural network.

OBJECTIVE: To develop deep learning-based cardiac chamber enlargement-detection algorithms for left atrial (DLCE-LAE) and ventricular enlargement (DLCE-LVE), on chest radiographs METHODS: For training and internal validation of DLCE-LAE and -LVE, 5,045 chest radiographs (CRs; 2,463 normal and 2,393 LAE) and 1,012 CRs (456 normal and 456 LVE) matched with the same-day echocardiography were collected, respectively. External validation was performed using 107 temporally independent CRs. Reader performance test was conducted using the external validation dataset by five cardiothoracic radiologists without and with the results of DLCE. Classification performance of DLCE was evaluated and compared with those of the readers and conventional radiographic features, including cardiothoracic ratio, carinal angle, and double contour. In addition, DLCE-LAE was tested on 5,277 CRs from a healthcare screening program cohort. RESULTS: DLCE-LAE showed areas under the receiver operating characteristics curve (AUROCs) of 0.858 on external validation. On reader performance test, DLCE-LAE showed better results than pooled radiologists (AUROC 0.858 vs. 0.651; p < .001) and significantly increased their performance when used as a second reader (AUROC 0.651 vs. 0.722; p < .001). DLCE-LAE also showed a significantly higher AUROC than conventional radiographic findings (AUROC 0.858 vs. 0.535-0.706; all ps < .01). In the healthcare screening cohort, DLCE-LAE successfully detected 71.0% (142/200) CRs with moderate-to-severe LAE (93.5% [29/31] of severe cases), while yielding 11.8% (492/4,184) false-positive rate. DLCE-LVE showed AUROCs of 0.966 and 0.594 on internal and external validation, respectively. CONCLUSION: DLCE-LAE outperformed and improved cardiothoracic radiologists' performance in detecting LAE and showed promise in screening individuals with moderate-to-severe LAE in a healthcare screening cohort. KEY POINTS: • Our deep learning algorithm outperformed cardiothoracic radiologists in detecting left atrial enlargement on chest radiographs. • Cardiothoracic radiologists improved their performance in detecting left atrial enlargement when aided by the algorithm. • On a healthcare-screening cohort, our algorithm detected 71.0% (142/200) radiographs with moderate-to-severe left atrial enlargement while yielding 11.8% (492/4,184) false-positive rate.

Weakly supervised lesion localization for age-related macular degeneration detection using optical coherence tomography images.

Age-related macular degeneration (AMD) is the main cause of irreversible blindness among the elderly and require early diagnosis to prevent vision loss, and careful treatment is essential. Optical coherence tomography (OCT), the most commonly used imaging method in the retinal area for the diagnosis of AMD, is usually interpreted by a clinician, and OCT can help diagnose disease on the basis of the relevant diagnostic criteria, but these judgments can be somewhat subjective. We propose an algorithm for the detection of AMD based on a weakly supervised convolutional neural network (CNN) model to support computer-aided diagnosis (CAD) system. Our main contributions are the following three things. (1) We propose a concise CNN model for OCT images, which outperforms the existing large CNN models using VGG16 and GoogLeNet architectures. (2) We propose an algorithm called Expressive Gradients (EG) that extends the existing Integrated Gradients (IG) algorithm so as to exploit not only the input-level attribution map, but also the high-level attribution maps. Due to enriched gradients, EG can highlight suspicious regions for diagnosis of AMD better than the guided-backpropagation method and IG. (3) Our method provides two visualization options: overlay and top-k bounding boxes, which would be useful for CAD. Through experimental evaluation using 10,100 clinical OCT images from AMD patients, we demonstrate that our EG algorithm outperforms the IG algorithm in terms of localization accuracy and also outperforms the existing object detection methods in terms of class accuracy.