A machine learning model to classify aortic dissection patients in the early diagnosis phase

A deep-learning system integrating electrocardiograms and laboratory indicators for diagnosing acute aortic dissection and acute myocardial infarction

BACKGROUND

Acute Stanford Type A aortic dissection (AAD-type A) and acute myocardial infarction (AMI) present with similar symptoms but require distinct treatments. Efficient differentiation is critical due to limited access to radiological equipment in many primary healthcare. This study develops a multimodal deep learning model integrating electrocardiogram (ECG) signals and laboratory indicators to enhance diagnostic accuracy for AAD-type A and AMI.

METHODS

We gathered ECG and laboratory data from 136 AAD-type A and 141 AMI patients at Zigong Fourth People's Hospital (January 2019 to December 2023) for training and validation. Utilizing ResNet-34 (residual network), we extracted ECG features and combined them with laboratory and demographic data. We assessed logistic regression, RandomForest, XGBoost, and LightGBM models, employing shapley additive explanations (SHAP) for feature importance analysis. Data from 30 AMI and 32 AAD-type A patients (January to September 2024) were used as a prospective test set.

RESULTS

Incorporating ECG features significantly improved model's AUC value, with the RandomForest achieving the best performance (AUC 0.98 on validation, 0.969 on test). SHAP analysis revealed that troponin and D-dimer, along with the embedding features of ECG extracted by the deep neural network, are key characteristics for differentiating AAD-type A and AMI.

CONCLUSION

ECG features are valuable for distinguishing AAD-type A and AMI, offering a novel tool for rapid cardiovascular disease diagnosis through multimodal data fusion and deep learning.

A deep learning algorithm for the detection of aortic dissection on non-contrast-enhanced computed tomography via the identification and segmentation of the true and false lumens of the aorta

Background

Aortic dissection is a life-threatening clinical emergency, but it is often missed and misdiagnosed due to the limitations of diagnostic technology. In this study, we developed a deep learning-based algorithm for identifying the true and false lumens in the aorta on non-contrast-enhanced computed tomography (NCE-CT) scans and to ascertain the presence of aortic dissection. Additionally, we compared the diagnostic performance of this algorithm with that of radiologists in detecting aortic dissection.

Methods

We included 320 patients with suspected acute aortic syndrome from three centers (Beijing Anzhen Hospital Affiliated to Capital Medical University, Fujian Provincial Hospital, and Xiangya Hospital of Central South University) between May 2020 and May 2022 in this retrospective study. All patients underwent simultaneous NCE-CT and contrast-enhanced CT (CE-CT). The cohort comprised 160 patients with aortic dissection and 160 without aortic dissection. A deep learning algorithm, three-dimensional (3D) full-resolution U-Net, was continuously trained and refined to segment the true and false lumens of the aorta to determine the presence of aortic dissection. The algorithm's efficacy in detecting dissections was evaluated using the receiver operating characteristic (ROC) curve, including the area under the curve (AUC), sensitivity, and specificity. Furthermore, a comparative analysis of the diagnostic capabilities between our algorithm and three radiologists was conducted.

Results

In diagnosing aortic dissection using NCE-CT images, the developed algorithm demonstrated an accuracy of 93.8% [95% confidence interval (CI): 89.8-98.3%], a sensitivity of 91.6% (95% CI: 86.7-95.8%), and a specificity of 95.6% (95% CI: 91.2-99.3%). In contrast, the radiologists achieved an accuracy of 88.8% (95% CI: 83.5-94.1%), a sensitivity of 90.6% (95% CI: 83.5-94.1%), and a specificity of 94.1% (95% CI: 72.9-97.6%). There was no significant difference between the algorithm's performance and radiologists' mean performance in accuracy, sensitivity, or specificity (P>0.05).

Conclusions

The algorithm proficiently segments the true and false lumens in aortic NCE-CT images, exhibiting diagnostic capabilities comparable to those of radiologists in detecting aortic dissection. This suggests that the algorithm could reduce misdiagnoses in clinical practice, thereby enhancing patient care.

Identifying Acute Aortic Syndrome and Thoracic Aortic Aneurysm from Chest Radiography in the Emergency Department Using Convolutional Neural Network Models

(1) Background: Identifying acute aortic syndrome (AAS) and thoracic aortic aneurysm (TAA) in busy emergency departments (EDs) is crucial due to their life-threatening nature, necessitating timely and accurate diagnosis. (2) Methods: This retrospective case-control study was conducted in the ED of three hospitals. Adult patients visiting the ED between 1 January 2010 and 1 January 2020 with a chief complaint of chest or back pain were enrolled in the study. The collected chest radiography (CXRs) data were divided into training (80%) and testing (20%) datasets. The training dataset was trained by four different convolutional neural network (CNN) models. (3) Results: A total of 1625 patients were enrolled in this study. The InceptionV3 model achieved the highest F1 score of 0.76. (4) Conclusions: Analysis of CXRs using a CNN-based model provides a novel tool for clinicians to interpret ED patients with chest pain and suspected AAS and TAA. The integration of such imaging tools into ED could be considered in the future to enhance the diagnostic workflow for clinically fatal diseases.

Image detection of aortic dissection complications based on multi-scale feature fusion

Background

Aortic dissection refers to the true and false two-lumen separation of the aortic wall, in which the blood in the aortic lumen enters the aortic mesomembrane from the tear of the aortic intima to separate the mesomembrane and expand along the long axis of the aorta.

Purpose

In view of the problems of individual differences, complex complications and many small targets in clinical aortic dissection detection, this paper proposes a convolution neural network MFF-FPN (Multi-scale Feature Fusion based Feature Pyramid Network) for the detection of aortic dissection complications.

Methods

The proposed model uses Resnet50 as the backbone for feature extraction and builds a pyramid structure to fuse low-level and high-level feature information. We add an attention mechanism to the backbone network, which can establish inter-dependencies between feature graph channels and enhance the representation quality of CNN.

Results

The proposed method has a mean average precision (MAP) of 99.40% in the task of multi object detection for aortic dissection and complications, which is higher than the accuracy of 96.3% on SSD model and 99.05% on YoloV7 model. It greatly improves the accuracy of small target detection such as cysts, making it more suitable for clinical focus detection.

Conclusions

The proposed deep learning model achieves feature reuse and focuses on local important information. By adding only a small number of model parameters, we are able to greatly improve the detection accuracy, which is effective in detecting small target lesions commonly found in clinical settings, and also performs well on other medical and natural datasets.

Enhancing patient outcomes: the role of clinical utility in guiding healthcare providers in curating radiology AI applications

With advancements in artificial intelligence (AI) dominating the headlines, diagnostic imaging radiology is no exception to the accelerating role that AI is playing in today's technology landscape. The number of AI-driven radiology diagnostic imaging applications (digital diagnostics) that are both commercially available and in-development is rapidly expanding as are the potential benefits these tools can deliver for patients and providers alike. Healthcare providers seeking to harness the potential benefits of digital diagnostics may consider evaluating these tools and their corresponding use cases in a systematic and structured manner to ensure optimal capital deployment, resource utilization, and, ultimately, patient outcomes-or clinical utility. We propose several guiding themes when using clinical utility to curate digital diagnostics.

Patterns in acute aortic dissection and a connection to meteorological conditions in Germany

Acute type A aortic dissection (ATAAD) is a dramatic emergency exhibiting a mortality of 50% within the first 48 hours if not operated. This study found an absolute value of cosine-like seasonal variation pattern for Germany with significantly fewer ATAAD events (Wilcoxon test) for the warm months of June, July, and August from 2005 to 2015. Many studies suspect a connection between ATAAD events and weather conditions. Using ERA5 reanalysis data and an objective weather type classification in a contingency table approach showed that for Germany, significantly more ATAAD events occurred during lower temperatures (by about 4.8 K), lower water vapor pressure (by about 2.6 hPa), and prevailing wind patterns from the northeast. In addition, we used data from a classification scheme for human-biometeorological weather conditions which was not used before in ATAAD studies. For the German region of Berlin and Brandenburg, for 2006 to 2019, the proportion of days with ATAAD events during weather conditions favoring hypertension (cold air advection, in the center of a cyclone, conditions with cold stress or thermal comfort) was significantly increased by 13% (Chi-squared test for difference of proportions). In contrast, the proportion was decreased by 19% for conditions associated with a higher risk for patients with hypotension and therefore a lower risk for patients with hypertension (warm air advection ahead of warm fronts, conditions with no thermal stress or heat stress, in the center of a cyclone with thermal stress). As many studies have shown that hypertension is a risk factor for ATAAD, our findings support the hypothesized relation between ATAAD and hypertension-favoring weather conditions.

Calibration and Distraction Mining Network for Aortic True Lumen segmentation

Automatic segmentation of aortic true lumen based on deep learning can save the time for diagnosis of aortic dissection. However, fuzzy boundary, small true lumen region, and high similarity usually leads to inaccurate prediction. To make better use of the details supplemented by the encoder to restore boundaries, we decompose the recovery of detail features in the decoder into two sub-processes: calibration and distraction mining. And we propose a novel calibration and distraction mining (CDM) module. It utilizes deep features to calibrate shallow features so that features are concentrated in the main region. Then, it leverages the distraction mining procedure to extract false-negative features as a supplement to calibrated features and recover details of the segmentation object. We construct CDM-Net and verify its performance on the Aorta-CT dataset (private dataset), it achieves the Dice similarity coefficient of 96.94% and the Jaccard index coefficient of 94.08%, which is the best compared with 10 latest methods. Similarly, we explore its robustness on three more public datasets, including ISIC 2018 dataset (skin lesion segmentation), the 2018 data science bowl dataset (nucleus segmentation), LUNA dataset (lung segmentation). Experimental results prove that our method produces competitive results on all three data sets. Through quantitative and qualitative research, the proposed CDM-Net has good performance and can process aortic slices with complex semantic features, additional experiments show that it has good robustness, and it has the potential to be applied and expanded conveniently.

Cluster-Based Ensemble Learning Model for Aortic Dissection Screening

Aortic dissection (AD) is a rare and high-risk cardiovascular disease with high mortality. Due to its complex and changeable clinical manifestations, it is easily missed or misdiagnosed. In this paper, we proposed an ensemble learning model based on clustering: Cluster Random under-sampling Smote–Tomek Bagging (CRST-Bagging) to help clinicians screen for AD patients in the early phase to save their lives. In this model, we propose the CRST method, which combines the advantages of Kmeans++ and the Smote–Tomek sampling method, to overcome an extremely imbalanced AD dataset. Then we used the Bagging algorithm to predict the AD patients. We collected AD patients’ and other cardiovascular patients’ routine examination data from Xiangya Hospital to build the AD dataset. The effectiveness of the CRST method in resampling was verified by experiments on the original AD dataset. Our model was compared with RUSBoost and SMOTEBagging on the original dataset and a test dataset. The results show that our model performed better. On the test dataset, our model’s precision and recall rates were 83.6% and 80.7%, respectively. Our model’s F1-score was 82.1%, which is 4.8% and 1.6% higher than that of RUSBoost and SMOTEBagging, which demonstrates our model’s effectiveness in AD screening.

Solving the class imbalance problem using ensemble algorithm: application of screening for aortic dissection

Background

Imbalance between positive and negative outcomes, a so-called class imbalance, is a problem generally found in medical data. Despite various studies, class imbalance has always been a difficult issue. The main objective of this study was to find an effective integrated approach to address the problems posed by class imbalance and to validate the method in an early screening model for a rare cardiovascular disease aortic dissection (AD).

Methods

Different data-level methods, cost-sensitive learning, and the bagging method were combined to solve the problem of low sensitivity caused by the imbalance of two classes of data. First, feature selection was applied to select the most relevant features using statistical analysis, including significance test and logistic regression. Then, we assigned two different misclassification cost values for two classes, constructed weak classifiers based on the support vector machine (SVM) model, and integrated the weak classifiers with undersampling and bagging methods to build the final strong classifier. Due to the rarity of AD, the data imbalance was particularly prominent. Therefore, we applied our method to the construction of an early screening model for AD disease. Clinical data of 523,213 patients from the Institute of Hypertension, Xiangya Hospital, Central South University were used to verify the validity of this method. In these data, the sample ratio of AD patients to non-AD patients was 1:65, and each sample contained 71 features.

Results

The proposed ensemble model achieved the highest sensitivity of 82.8%, with training time and specificity reaching 56.4 s and 71.9% respectively. Additionally, it obtained a small variance of sensitivity of 19.58 × 10^–3 in the seven-fold cross validation experiment. The results outperformed the common ensemble algorithms of AdaBoost, EasyEnsemble, and Random Forest (RF) as well as the single machine learning (ML) methods of logistic regression, decision tree, k nearest neighbors (KNN), back propagation neural network (BP) and SVM. Among the five single ML algorithms, the SVM model after cost-sensitive learning method performed best with a sensitivity of 79.5% and a specificity of 73.4%.

Conclusions

In this study, we demonstrate that the integration of feature selection, undersampling, cost-sensitive learning and bagging methods can overcome the challenge of class imbalance in a medical dataset and develop a practical screening model for AD, which could lead to a decision support for screening for AD at an early stage.

A Combined Deep Learning System for Automatic Detection of “Bovine” Aortic Arch on Computed Tomography Scans

The “bovine” aortic arch is an anatomic variant consisting in a common origin of the innominate and left carotid artery (CILCA), associated with a greater risk of thoracic aortic diseases (aneurysms and dissections), stroke, and complications after endovascular procedures. CILCA can be detected by visual assessment of computed tomography (CT) chest scans, but it is rarely reported. We developed a deep learning (DL) segmentation-plus-classification system to automatically detect CILCA based on 302 CT studies acquired at 2 centers. One model (3D U-Net) was trained from scratch (supervised by manual segmentation), validated, and tested for the automatic segmentation of the aortic arch and supra-aortic vessels. Three DL architectures (ResNet50, DenseNet-201, and SqueezeNet), pre-trained over millions of common images, were trained, validated, and tested for the automatic classification of CILCA versus non-CILCA, supervised by radiologist’s classification. The 3D U-Net-plus-DenseNet-201 was found to be the best system (Dice index 0.912); its classification performance obtained from internal, independent testing on 126 patients gave a receiver operating characteristic area under the curve of 87.0%, sensitivity 66.7%, specificity 90.5%, positive predictive value 87.5%, negative predictive value 73.1%, positive likelihood ratio 7.0, and negative likelihood ratio 0.4. In conclusion, a combined DL system applied to chest CT scans was developed and proven to be an effective tool to detect individuals with “bovine” aortic arch with a low rate of false-positive findings.

Aortic Dissection Auxiliary Diagnosis Model and Applied Research Based on Ensemble Learning

Aortic dissection (AD), a dangerous disease threatening to human beings, has a hidden onset and rapid progression and has few effective methods in its early diagnosis. At present, although CT angiography acts as the gold standard on AD diagnosis, it is so expensive and time-consuming that it can hardly offer practical help to patients. Meanwhile, the artificial intelligence technology may provide a cheap but effective approach to building an auxiliary diagnosis model for improving the early AD diagnosis rate by taking advantage of the data of the general conditions of AD patients, such as the data about the basic inspection information. Therefore, this study proposes to hybrid five types of machine learning operators into an integrated diagnosis model, as an auxiliary diagnostic approach, to cooperate with the AD-clinical analysis. To improve the diagnose accuracy, the participating rate of each operator in the proposed model may adjust adaptively according to the result of the data learning. After a set of experimental evaluations, the proposed model, acting as the preliminary AD-discriminant, has reached an accuracy of over 80%, which provides a promising instance for medical colleagues.

DRG grouping by machine learning: from expert-oriented to data-based method

Background

Diagnosis-related groups (DRGs) are a payment system that could effectively solve the problem of excessive increases in healthcare costs which are applied as a principal measure in the healthcare reform in China. However, expert-oriented DRG grouping is a black box with the drawbacks of upcoding and high cost.

Methods

This study proposes a method of data-based grouping, designed and updated by machine learning algorithms, which could be trained by real cases, or even simulated cases. It inherits the decision-making rules from the expert-oriented grouping and improves performance by incorporating continuous updates at low cost. Five typical classification algorithms were assessed and some suggestions were made for algorithm choice. The kappa coefficients were reported to evaluate the performance of grouping.

Results

Based on tenfold cross-validation, experiments showed that data-based grouping had a similar classification performance to the expert-oriented grouping when choosing suitable algorithms. The groupings trained by simulated cases had less accuracy when they were tested by the real cases rather than simulated cases, but the kappa coefficients of the best model were still higher than 0.6. When the grouping was tested in a new DRGs system, the average kappa coefficients were significantly improved from 0.1534 to 0.6435 by the update; and with enough computation resources, the update process could be completed in a very short time.

Conclusions

As a new potential option, the data-based grouping meets the requirements of the DRGs system and has the advantages of high transparency and low cost in the design and update process.

Machine Learning Models for Predicting In-Hospital Mortality in Acute Aortic Dissection Patients

Background: Acute aortic dissection is a potentially fatal cardiovascular disorder associated with high mortality. However, current predictive models show a limited ability to efficiently and flexibly detect this mortality risk, and have been unable to discover a relationship between the mortality rate and certain variables. Thus, this study takes an artificial intelligence approach, whereby clinical data-driven machine learning was utilized to predict the in-hospital mortality of acute aortic dissection. Methods: Patients diagnosed with acute aortic dissection between January 2015 to December 2018 were voluntarily enrolled from the Second Xiangya Hospital of Central South University in the study. The diagnosis was defined by magnetic resonance angiography or computed tomography angiography, with an onset time of the symptoms being within 14 days. The analytical variables included demographic characteristics, physical examination, symptoms, clinical condition, laboratory results, and treatment strategies. The machine learning algorithms included logistic regression, decision tree, K nearest neighbor, Gaussian naive bayes, and extreme gradient boost (XGBoost). Evaluation of the predictive performance of the models was mainly achieved using the area under the receiver operating characteristic curve. SHapley Additive exPlanation was also implemented to interpret the final prediction model. Results: A total of 1,344 acute aortic dissection patients were recruited, including 1,071 (79.7%) patients in the survivor group and 273 (20.3%) patients in non-survivor group. The extreme gradient boost model was found to be the most effective model with the greatest area under the receiver operating characteristic curve (0.927, 95% CI: 0.860-0.968). The three most significant aspects of the extreme gradient boost importance matrix plot were treatment, type of acute aortic dissection, and ischemia-modified albumin levels. In the SHapley Additive exPlanation summary plot, medical treatment, type A acute aortic dissection, and higher ischemia-modified albumin level were shown to increase the risk of hospital-based mortality.

A Deep-Learning Algorithm-Enhanced System Integrating Electrocardiograms and Chest X-rays for Diagnosing Aortic Dissection

BACKGROUND

Chest pain is the most common symptom of aortic dissection (AD), but it is often confused with other prevalent cardiopulmonary diseases. We aimed to develop deep-learning models (DLMs) with electrocardiography (ECG) and chest x-ray (CXR) features to detect AD and evaluate their performance.

METHODS

This study included 43,473 patients in the emergency department (ED) between July 2012 and December 2019 for retrospective DLM development. A development cohort including 49,071 ED records (120 AD type A and 64 AD type B) was used to train DLMs for ECG and CXR, and 9904 independent ED records (40 AD type A and 34 AD type B) were used to validate DLM performance. Human-machine competitions of ECG and CXR were conducted. Patient characteristics and laboratory results were used to enhance the diagnostic accuracy. The DLM-enabled AD diagnostic process was prospectively evaluated in 25,885 ED visits.

RESULTS

The area under the curves (AUCs) of the ECG and CXR models were 0.918 and 0.857 for detecting AD in a human-machine competition, respectively, which were better than those of the participating physicians. In the validation cohort, the AUCs of the integrated model were 0.882, 0.960, and 0.813 in all AD, AD type A, and AD type B patients, respectively, with a sensitivity of 100.0% and a specificity of 81.7% for AD type A. In patients with chest pain and D-dimer tests, the DLM could predict more precisely, achieving a positive predictive value of 62.5% in the prospective evaluation.

CONCLUSIONS

DLMs may serve as decision-supporting tools for identification of AD and facilitate differential diagnosis in patients with acute chest pain.

An early aortic dissection screening model and applied research based on ensemble learning

Background

As a particularly dangerous and rare cardiovascular disease, aortic dissection (AD) is characterized by complex and diverse symptoms and signs. In the early stage, the rate of misdiagnosis and missed diagnosis is relatively high. This study aimed to use machine learning technology to establish a fast and accurate screening model that requires only patients' routine examination data as input to obtain predictive results.

Methods

A retrospective analysis of the examination data and diagnosis results of 53,213 patients with cardiovascular disease was conducted. Among these samples, 802 samples had AD. Forty-two features were extracted from the patients' routine examination data to establish a prediction model. There were five ensemble learning models applied to explore the possibility of using machine learning methods to build screening models for AD, including AdaBoost, XGBoost, SmoteBagging, EasyEnsemble and XGBF. Among these, XGBF is an ensemble learning model that we propose to deal with the imbalance of the positive and negative samples. The seven-fold cross validation method was used to analyze and verify the performance of each model. Due to the imbalance of the samples, the evaluation indicators were sensitivity and specificity.

Results

Comparative experiments showed that the sensitivity of XGBF was 80.5%, which was better than the 16.1% of AdaBoost, 15.7% of XGBoost, 78.0% of SmoteBagging and 77.8% of EasyEnsemble. Additionally, XGBF had relatively high specificity, and the training time consumption was short. Based on these three indicators, XGBF performed best, and met the application requirements, which means through careful design, we can use machine learning technology to achieve early AD screening.

Conclusions

Through reasonable design, the ensemble learning method can be used to build an effective screening model. The XGBF has high practical application value for screening for AD.

A study of aortic dissection screening method based on multiple machine learning models

Background

The main purpose of the study was to develop an early screening method for aortic dissection (AD) based on machine learning. Due to the rarity of AD and the complexity of symptoms, many doctors have no clinical experience with it. Many patients are not suspected of having AD, which lead to a high rate of misdiagnosis. Here, we report the preliminary study and feasibility of rapid and accurate screening method of AD with machine learning methods.

Methods

The dataset analyzed was composed by examination data provided by the Xiangya Hospital Central South University of China which include a total of 60,000 samples, including aortic patients and non-aortic ones. Each sample has 76 features which is consist of routine examinations and other easily accessible information. Since the proportion of people who are affected is usually imbalanced compared to non-diseased people, multiple machine learning models were used, include AdaBoost, SmoteBagging, EasyEnsemble and CalibratedAdaMEC. They used different methods such as ensemble learning, undersampling, oversampling, and cost-sensitivity to solve data imbalance problems.

Results

AdaBoost performed poorly with an average recall of 16.1% and a specificity of 99.8%. SmoteBagging achieved a statistically significant better performance for this problem with an average recall of 78.1% and a specificity of 79.2%. EasyEnsemble reached the values of 77.8% and 79.3% for recall and specificity respectively. CalibratedAdaMEC’s recall and specificity are 75.8% and 76%.

Conclusions

It was found that the screening performance of the models evaluated in this paper had a misdiagnosis rate lower than 25% except AdaBoost. The data used in these methods are only routine inspection data. This means that machine learning methods can help us build a fast, cheap, worthwhile and effective early screening approach for AD.