Segmentation-based cardiomegaly detection based on semi-supervised estimation of cardiothoracic ratio.

The successful integration of neural networks in a clinical setting is still uncommon despite major successes achieved by artificial intelligence in other domains. This is mainly due to the black box characteristic of most optimized models and the undetermined generalization ability of the trained architectures. The current work tackles both issues in the radiology domain by focusing on developing an effective and interpretable cardiomegaly detection architecture based on segmentation models. The architecture consists of two distinct neural networks performing the segmentation of both cardiac and thoracic areas of a radiograph. The respective segmentation outputs are subsequently used to estimate the cardiothoracic ratio, and the corresponding radiograph is classified as a case of cardiomegaly based on a given threshold. Due to the scarcity of pixel-level labeled chest radiographs, both segmentation models are optimized in a semi-supervised manner. This results in a significant reduction in the costs of manual annotation. The resulting segmentation outputs significantly improve the interpretability of the architecture's final classification results. The generalization ability of the architecture is assessed in a cross-domain setting. The assessment shows the effectiveness of the semi-supervised optimization of the segmentation models and the robustness of the ensuing classification architecture.

Stretch-induced compliance mechanism in pregnancy-induced cardiac hypertrophy and the impact of cardiovascular risk factors.

Pressure overload-induced hypertrophy compromises cardiac stretch-induced compliance (SIC) after acute volume overload (AVO). We hypothesized that SIC could be enhanced by physiological hypertrophy induced by pregnancy's chronic volume overload. This study evaluated SIC-cardiac adaptation in pregnant women with or without cardiovascular risk (CVR) factors. Thirty-seven women (1st trimester, 1stT) and a separate group of 31 (3rd trimester, 3rdT) women [healthy or with CVR factors (obesity and/or hypertension and/or with gestational diabetes)] underwent echocardiography determination of left ventricular end-diastolic volume (LVEDV) and E/e' before (T0), immediately after (T1), and 15 min after (T2; SIC) AVO induced by passive leg elevation. Blood samples for NT-proBNP quantification were collected before and after the AVO. Acute leg elevation significantly increased inferior vena cava diameter and stroke volume from T0 to T1 in both 1stT and 3rdT, confirming AVO. LVEDV and E/e' also increased immediately after AVO (T1) in both 1stT and 3rdT. SIC adaptation (T2, 15 min after AVO) significantly decreased E/e' in both trimesters, with additional expansion of LVEDV only in the 1stT. NT-pro-BNP increased slightly after AVO but only in the 1stT. CVR factors, but not parity or age, significantly impacted SIC cardiac adaptation. A distinct functional response to SIC was observed between 1stT and 3rdT, which was influenced by CVR factors. The LV of 3rdT pregnant women was hypertrophied, showing a structural limitation to dilate with AVO, whereas the lower LV filling pressure values suggest increased diastolic compliance.NEW & NOTEWORTHY The sudden increase of volume overload triggers an acute myocardial stretch characterized by an immediate rise in contractility by the Frank-Starling mechanism, followed by a progressive increase known as the slow force response. The present study is the first to characterize echocardiographically the stretch-induced compliance (SIC) mechanism in the context of physiological hypertrophy induced by pregnancy. A distinct functional adaptation to SIC was observed between first and third trimesters, which was influenced by cardiovascular risk factors.

A deep-learning-based framework for identifying and localizing multiple abnormalities and assessing cardiomegaly in chest X-ray.

Accurate identification and localization of multiple abnormalities are crucial steps in the interpretation of chest X-rays (CXRs); however, the lack of a large CXR dataset with bounding boxes severely constrains accurate localization research based on deep learning. We created a large CXR dataset named CXR-AL14, containing 165,988 CXRs and 253,844 bounding boxes. On the basis of this dataset, a deep-learning-based framework was developed to identify and localize 14 common abnormalities and calculate the cardiothoracic ratio (CTR) simultaneously. The mean average precision values obtained by the model for 14 abnormalities reached 0.572-0.631 with an intersection-over-union threshold of 0.5, and the intraclass correlation coefficient of the CTR algorithm exceeded 0.95 on the held-out, multicentre and prospective test datasets. This framework shows an excellent performance, good generalization ability and strong clinical applicability, which is superior to senior radiologists and suitable for routine clinical settings.

Regressive vision transformer for dog cardiomegaly assessment.

Cardiac disease is one of the leading causes of death in dogs. Automatic cardiomegaly detection has great significance in helping clinicians improve the accuracy of the diagnosis process. Deep learning methods show promising results in improving cardiomegaly classification accuracy, while they are still not widely applied in clinical trials due to the difficulty in mapping predicted results with input radiographs. To overcome these challenges, we first collect large-scale dog heart X-ray images. We then develop a dog heart labeling tool and apply a few-shot generalization strategy to accelerate the label speed. We also develop a regressive vision transformer model with an orthogonal layer to bridge traditional clinically used VHS metric with deep learning models. Extensive experimental results demonstrate that the proposed model achieves state-of-the-art performance.

Right Atrial Enlargement on Electrocardiogram in Previously Healthy Young Patients.

OBJECTIVES: To assess whether right atrial enlargement (RAE) on electrocardiogram (ECG) correlates with true RAE on echocardiogram in previously healthy young patients and to understand which patients with RAE on ECG may warrant additional testing. STUDY DESIGN: A single-center, retrospective review of previously healthy young patients with (1) ECGs that were read as RAE by a pediatric cardiologist and (2) echocardiograms obtained within 90 days of the ECG. ECGs were reviewed to confirm RAE and determine which leads met criteria. The echocardiograms were then reviewed and RA measurements with z scores obtained. A z score >2 was considered positive for RAE on echocardiogram. RESULTS: In total, 162 patients with median age 10.8 years were included in the study. A total of 23 patients had true RAE on echocardiogram, giving a positive predictive value (PPV) of 14%. In patients <1 year of age, the PPV increased to 35%. In patients older than 1 year, the PPV was low at 7%. Patients with true RAE were more likely to meet criteria for RAE in the anterior precordial leads (V1-V3) (48% vs 5%, P < .001) and meet criteria for right ventricular hypertrophy (22% vs 6%, P = .023). CONCLUSION: Our findings show that RAE on ECG has a low PPV for RAE on echocardiogram in previously healthy young patients. The highest yield for RAE on echocardiogram was observed in patients who were <1 year of age, had RAE in the anterior precordial leads, or displayed right ventricular hypertrophy on ECG.

Prognostic dilemmas for SIDS in idiopathic fetal right atrium dilatation: Case report and review literature.

The authors describe a case of fetal isolated right atrial enlargement or IDRA (idiopathic dilatations of the right atrium) evident in third trimester, complicated by arrhythmia in the female infant during the 1° month of life with ECG diagnosis of Wolf-Parkinson-White syndrome (WPW). The eldest sister died at 6 years because of an arrhythmia with the same diagnosis of WPW. The review of the literature on IDRA frequently shows a familial genetic aggregation. The pathogenetic mechanism underlying the dilation of the right atrium could consist of a myopathy or electrical conduction disorder. The exclusive involvement of the right atrium may be due to the increased pressure in the fetal right atrium. On the basis of our case and after review of the literature, we must be careful in defining as physiological the enlargement of the right fetal atrium in the third trimester of pregnancy. The ultrasound sign of IDRA may be a fetal prodrome of SIDS (sudden infant death syndrome).

Utility of vertebral left atrial size and vertebral heart size to aid detection of congestive heart failure in dogs with respiratory signs.

BACKGROUND: Differentiating cardiogenic vs noncardiogenic causes of respiratory signs can be challenging when echocardiography is unavailable. Radiographic vertebral left atrial size (VLAS) and vertebral heart size (VHS) have been shown to predict echocardiographic left heart size, with VLAS specifically estimating left atrial size. HYPOTHESIS/OBJECTIVES: Compare the diagnostic accuracy of VLAS and VHS to predict left-sided congestive heart failure (CHF) in dogs presenting with respiratory signs. ANIMALS: One-hundred fourteen dogs with respiratory signs and radiographic pulmonary abnormalities. METHODS: Retrospective cross-sectional study. Dogs had to have an echocardiogram and thoracic radiographs obtained within 24 hours. Diagnosis of CHF was confirmed based on the presence of respiratory signs, cardiac disease, LA enlargement, and cardiogenic pulmonary edema. RESULTS: Fifty-seven dogs had CHF and 57 did not have CHF. Compared to VHS (area under the curve [AUC] 0.85; 95% confidence interval [CI], 0.77-0.91), VLAS was a significantly (P = .03) more accurate predictor of CHF (AUC, 0.92; 95% CI, 0.85-0.96). Optimal cutoff for VLAS was >2.3 vertebrae (sensitivity, 93.0%; specificity, 82.5%). Murmur grade (P = .02) and VLAS (P < .0001) were independently associated with CHF and VHS was not. Increased VHS (54%) was significantly (P = .01) more common than increased VLAS (24%) in dogs without CHF. Results were similar in a subsample of older and smaller dogs. CONCLUSIONS AND CLINICAL IMPORTANCE: When echocardiography is unavailable, VLAS and murmur grade have clinically utility to aid in differentiating cardiogenic from noncardiogenic respiratory signs. These findings might be especially useful to help rule out CHF in dogs with increased VHS that present with respiratory signs.

Measuring cardiomyocyte cellular characteristics in cardiac hypertrophy using diffusion-weighted MRI.

PURPOSE: This paper presents a hierarchical modeling approach for estimating cardiomyocyte major and minor diameters and intracellular volume fraction (ICV) using diffusion-weighted MRI (DWI) data in ex vivo mouse hearts. METHODS: DWI data were acquired on two healthy controls and two hearts 3 weeks post transverse aortic constriction (TAC) using a bespoke diffusion scheme with multiple diffusion times ( Δ $$ \Delta $$ ), q-shells and diffusion encoding directions. Firstly, a bi-exponential tensor model was fitted separately at each diffusion time to disentangle the dependence on diffusion times from diffusion weightings, that is, b-values. The slow-diffusing component was attributed to the restricted diffusion inside cardiomyocytes. ICV was then extrapolated at Δ = 0 $$ \Delta =0 $$ using linear regression. Secondly, given the secondary and the tertiary diffusion eigenvalue measurements for the slow-diffusing component obtained at different diffusion times, major and minor diameters were estimated assuming a cylinder model with an elliptical cross-section (ECS). High-resolution three-dimensional synchrotron X-ray imaging (SRI) data from the same specimen was utilized to evaluate the biophysical parameters. RESULTS: Estimated parameters using DWI data were (control 1/control 2 vs. TAC 1/TAC 2): major diameter-17.4 µ $$ \mu $$ m/18.0 µ $$ \mu $$ m versus 19.2 µ $$ \mu $$ m/19.0 µ $$ \mu $$ m; minor diameter-10.2 µ $$ \mu $$ m/9.4 µ $$ \mu $$ m versus 12.8 µ $$ \mu $$ m/13.4 µ $$ \mu $$ m; and ICV-62%/62% versus 68%/47%. These findings were consistent with SRI measurements. CONCLUSION: The proposed method allowed for accurate estimation of biophysical parameters suggesting cardiomyocyte diameters as sensitive biomarkers of hypertrophy in the heart.

A convolutional attention mapping deep neural network for classification and localization of cardiomegaly on chest X-rays.

Building a reliable and precise model for disease classification and identifying abnormal sites can provide physicians assistance in their decision-making process. Deep learning based image analysis is a promising technique for enriching the decision making process, and accordingly strengthening patient care. This work presents a convolutional attention mapping deep learning model, Cardio-XAttentionNet, to classify and localize cardiomegaly effectively. We revisit the global average pooling (GAP) system and add a weighting term to develop a light and effective Attention Mapping Mechanism (AMM). The model enables the classification of cardiomegaly from chest X-rays through image-level classification and pixel-level localization only from image-level labels. We leverage some of the advanced ConvNet architectures as a backbone-model of the proposed attention mapping network to build Cardio-XAttentionNet. The proposed model is trained on ChestX-Ray14, which is a publicly accessible chest X-ray dataset. The best single model achieves an overall precision, recall, F-1 measure and area under curve (AUC) scores of 0.87, 0.85, 0.86 and 0.89, respectively, for the classification of the cardiomegaly. The results also demonstrate that the Cardio-XAttentionNet model well captures the cardiomegaly class information at image-level as well as localization at pixel-level on chest x-rays. A comparative analysis between the proposed AMM and existing GAP based models shows that the proposed model achieves a state-of-the-art performance on this dataset for cardiomegaly detection using a single model.

Cardiac hypertrophy simulations using parametric and echocardiography-based left ventricle model with shell finite elements.

In our paper, we simulated cardiac hypertrophy with the use of shell elements in parametric and echocardiography-based left ventricle (LV) models. The hypertrophy has an impact on the change in the wall thickness, displacement field and the overall functioning of the heart. We computed both eccentric and concentric hypertrophy effects and tracked changes in the ventricle shape and wall thickness. Thickening of the wall was developed under the influence of concentric hypertrophy, while the eccentric hypertrophy produces wall thinning. To model passive stresses we used the recently developed material modal based on the Holzapfel experiments. Also, our specific shell composite finite element models for heart mechanics are much smaller and simpler to use with respect to conventional 3D models. Furthermore, the presented modeling approach of the echocardiography-based LV can serve as the basis for practical applications since it relies on the true patient-specific geometry and experimental constitutive relationships. Our model gives an insight into hypertrophy development in realistic heart geometries, and it has the potential to test medical hypotheses regarding hypertrophy evolution in a healthy and heart with a disease, under the influence of different conditions and parameters.

A morphological study comparing heart size measurement on chest radiograph with echocardiography in local population.

OBJECTIVE: To investigate heart size on chest X-ray via cardiothoracic ratio, and to correlate it with echocardiographic measurements. METHODS: The comparative analytical, cross-sectional study was conducted at the Pakistan Navy Station Shifa Hospital, Karachi, between January 2021 and July 2021. The radiological parameters were measured on chest X-rays posterior-anterior view, and the echocardiographic parameters were measured using 2-dimensional transthoracic echocardiography. The absence or presence of cardiomegaly on both imaging modalities was modelled as a binary categorical variable and compared. Data was analysed using SPSS 23. RESULTS: Of the 79 participants, 44(55.7%) were males and 35(44.3%) were females. The mean age of the sample was 52.71±14.54 years. There were 28(35.44%) enlarged hearts on chest X-ray and 46(58.22%) on echocardiography. The sensitivity and specificity of chest X-ray were 54.35% and 90.90%, respectively. The positive and negative predictive values were 89.28% and 58.82%, respectively. The accuracy of chest X-ray in identifying an enlarged heart was 69.62%. CONCLUSIONS: The cardiac silhouette on a chest X-ray could demonstrate heart size through simple measurements with high specificity and reasonable accuracy. However, a normal heart size on chest X-ray may not have a normal function.

Accuracy of the modified vertebral heart score and the cardio-vertebral ratio for radiographic evaluation of cardiomegaly in ferrets.

Thoracic radiography is commonly used to assess the size of the heart and diagnose cardiac disease in ferrets. Several standardized radiographic heart size indicators have been introduced in this species and values in healthy ferrets have been reported. To date, none of these indicators has been tested in ferrets with cardiac disease. The aim of this prospective and retrospective, analytical observational design study was to assess the accuracy of the modified vertebral heart score (mVHS) and the cardio-vertebral ratio (CVR) in the radiographic detection of cardiomegaly in ferrets. Thoracic radiographs of 24 ferrets with confirmed heart diseases, 22 ferrets with non-cardiac diseases and normal-sized hearts on echocardiogram, and 24 healthy ferrets were mixed and examined by three independent and blinded radiologists who measured mVHS and CVR in right lateral (RL) and ventrodorsal (VD) radiographs. For all readers, ferrets with cardiac disease had significantly higher mVHS and CVR than ferrets without cardiac disease on echocardiography. Optimal cut-points for predicting cardiac enlargement were 6.25 vertebrae and 7.25 vertebrae for RL-mVHS and VD-mVHS, and 1.58 and 1.80 for RL-CVR and VD-CVR, respectively. Using these cut-points, the accuracy was good for indicators measured in RL radiographs (92.9% for RL-mVHS; 91.4% for RL-CVR) and moderate for indicators measured in VD radiographs (88.6% for VD-mVHS; 85.7% for VD-CVR). Findings supported the use of mVHS and CVR for evaluating the size of the heart in diseased ferrets, with caution in values interpretation when pericardial fat prevents precise delineation of the cardiac silhouette contour especially on VD radiographs.

SegNet-based left ventricular MRI segmentation for the diagnosis of cardiac hypertrophy and myocardial infarction.

OBJECTIVE: A set of cardiac MRI short-axis image dataset is constructed, and an automatic segmentation based on an improved SegNet model is developed to evaluate its performance based on deep learning techniques. METHODS: The Affiliated Hospital of Qingdao University collected 1354 cardiac MRI between 2019 and 2022, and the dataset was divided into four categories: for the diagnosis of cardiac hypertrophy and myocardial infraction and normal control group by manual annotation to establish a cardiac MRI library. On the basis, the training set, validation set and test set were separated. SegNet is a classical deep learning segmentation network, which borrows part of the classical convolutional neural network, that pixelates the region of an object in an image division of levels. Its implementation consists of a convolutional neural network. Aiming at the problems of low accuracy and poor generalization ability of current deep learning frameworks in medical image segmentation, this paper proposes a semantic segmentation method based on deep separable convolutional network to improve the SegNet model, and trains the data set. Tensorflow framework was used to train the model and the experiment detection achieves good results. RESULTS: In the validation experiment, the sensitivity and specificity of the improved SegNet model in the segmentation of left ventricular MRI were 0.889, 0.965, Dice coefficient was 0.878, Jaccard coefficient was 0.955, and Hausdorff distance was 10.163 mm, showing good segmentation effect. CONCLUSION: The segmentation accuracy of the deep learning model developed in this paper can meet the requirements of most clinical medicine applications, and provides technical support for left ventricular identification in cardiac MRI.

Development of non-bias phenotypic drug screening for cardiomyocyte hypertrophy by image segmentation using deep learning.

The number of patients with heart failure and related deaths is rapidly increasing worldwide, making it a major problem. Cardiac hypertrophy is a crucial preliminary step in heart failure, but its treatment has not yet been fully successful. In this study, we established a system to evaluate cardiomyocyte hypertrophy using a deep learning-based high-throughput screening system and identified drugs that inhibit it. First, primary cultured cardiomyocytes from neonatal rats were stimulated by both angiotensin II and endothelin-1, and cellular images were captured using a phase-contrast microscope. Subsequently, we used a deep learning model for instance segmentation and established a system to automatically and unbiasedly evaluate the cardiomyocyte size and perimeter. Using this system, we screened 100 FDA-approved drugs library and identified 12 drugs that inhibited cardiomyocyte hypertrophy. We focused on ezetimibe, a cholesterol absorption inhibitor, that inhibited cardiomyocyte hypertrophy in a dose-dependent manner in vitro. Additionally, ezetimibe improved the cardiac dysfunction induced by pressure overload in mice. These results suggest that the deep learning-based system is useful for the evaluation of cardiomyocyte hypertrophy and drug screening, leading to the development of new treatments for heart failure.

O-Ring Aortic Banding Versus Traditional Transverse Aortic Constriction for Modeling Pressure Overload-Induced Cardiac Hypertrophy.

Aortic banding in mice is one of the most commonly used experimental models for cardiac pressure overload-induced cardiac hypertrophy and the induction of heart failure. The previously used technique is based on a threaded suture around the aortic arch tied over a blunted 27 G needle to create stenosis. This method depends on the surgeon manually tightening the thread and, thus, leads to high variance in the diameter size. A newly refined method described by Melleby et al. promises less variance and more reproducibility after surgery. The new technique, o-ring- aortic banding (ORAB), uses a non-slip rubber ring instead of a suture with a thread, resulting in reduced variation in pressure overload and reproducible phenotypes of cardiac hypertrophy. During surgery, the o-ring is placed between the brachiocephalic and left carotid arteries. Successful constriction is confirmed by echocardiography. After 1 day, correct placement of the ring results in an increased flow velocity in the transverse aorta over the o-ring-induced stenosis. After 2 weeks, impaired cardiac function is proven by decreased ejection fraction and increased wall thickness. Importantly, besides less variance in the diameter size, ORAB is associated with lower intra- and post-operative mortality rates compared with transverse aortic constriction (TAC). Thus, ORAB represents a superior method to the commonly used TAC surgery, resulting in more reproducible results and a possible reduction in the number of animals needed.