KEYWORDS: Image segmentation, Aorta, Magnetic resonance imaging, 3D modeling, Hemodynamics, Electroluminescence, Blood circulation, Deep learning, Education and training, Data modeling
PurposeTo develop an automated method for aortic segmentation using deep learning techniques and further analyze the hemodynamic parameters in patients with bicuspid aortic valve (BAV). Since four-dimensional (4D) flow magnetic resonance imaging (MRI) imaging helps in analyzing and quantifying the blood flow changes that occur in aortic valve-related problems, such as BAV, 4D flow MRI images are considered.ApproachOur dataset consisted of 91 patients who had referral indications of BAV and 30 healthy volunteers who had no known cardiovascular disease. A U-Net++ with pretrained ResNet-34 encoders was trained for aortic segmentation using manual segmentation by an expert as the ground truth. In the first stage, the model was evaluated on 21 test cohorts using overlay and distance-based metrics, such as Dice score, Hausdorff distance, and absolute volume difference. In the second stage, the hemodynamic parameters, such as wall shear stress (WSS), viscous energy loss, and vorticity, were calculated to quantify the blood flow irregularities that occur in BAV patients. The segmentation and the flow parameters generated by the algorithm were compared with those generated using the manual segmentations. Paired t-test with alpha value of 0.05 was used for statistical significance testing.ResultsAs for overlap and distance-based metrics, the developed algorithm reported a Dice score coefficient of 0.90 ± 0.03, absolute volume difference of 1683 ± 1139 mm3, and Hausdorff distance of 3.2 ± 1.18 mm on test cohorts. The hemodynamic parameters calculated between automated and manual methods resulted in a mean difference of 6.62% for WSS with p-value of 0.94, 17.35% for mean viscous energy loss with p-value of 0.78, and 7.59% for vorticity with p-value of 0.97.ConclusionsA fast and accurate segmentation tool was developed for aortic segmentation using a dataset taken at clinical and blood flow parameters that were calculated based on the segmented aorta. These results will assist the clinicians to analyze the blood flow patterns and commence distinguished treatment in BAV patients.
Bicuspid aortic valve (BAV) is a hereditary disorder that develops in the fetus at the early stages of pregnancy. Though the patient may have BAV defect at the time of birth, it may not be diagnosed until the patient becomes often symptomatic in adulthood. BAV patients are at a higher risk of aneurysm growth with a high mortality rate. Hence, measurements acquired from automated aortic segmentation would aid in faster analysis of hemodynamic parameters for better risk-stratify in BAV patients. In this work, we propose a fully automated segmentation tool using a deep learning technique for fast and accurate aortic segmentation. The 3D aorta volume was segmented based on the proposed model (U-Net++) and compared with two-dimensional (2D) deep convolutional neural network (DCNN) models (U-Net and Attention U-Net). Performance metrics such as Dice similarity coefficient (DSC), Hausdorff distance (HD), and absolute volume difference (AVD) were used for model evaluation. The proposed model reported the highest DSC of 0.88±0.02 on the dataset comprising of 114 subjects (n=91 BAV and n=23 healthy cases). The HD shows a difference in mean of 3.8mm between the manual and the predicted results. Though a limited dataset was deployed in this work, the model reports a high DSC based on 3D phase contrast (PC) magnetic resonance angiogram (MRA) (PCMRA) images obtained at a clinical setting. This fully automated approach minimizes the burdensome data analysis, data annotation cost and would aid for early diagnosis and to start individualized treatment to enhance the patient outcome.
Purpose: The objective of this study is to develop and evaluate a fully automated, deep learning-based method for detection of COVID-19 infection from chest x-ray images.
Approach: The proposed model was developed by replacing the final classifier layer in DenseNet201 with a new network consisting of global averaging layer, batch normalization layer, a dense layer with ReLU activation, and a final classification layer. Then, we performed an end-to-end training using the initial pretrained weights on all the layers. Our model was trained using a total of 8644 images with 4000 images each in normal and pneumonia cases and 644 in COVID-19 cases representing a large real dataset. The proposed method was evaluated based on accuracy, sensitivity, specificity, ROC curve, and F1-score using a test dataset comprising 1729 images (129 COVID-19, 800 normal, and 800 pneumonia). As a benchmark, we also compared the results of our method with those of seven state-of-the-art pretrained models and with a lightweight CNN architecture designed from scratch.
Results: The proposed model based on DenseNet201 was able to achieve an accuracy of 94% in detecting COVID-19 and an overall accuracy of 92.19%. The model was able to achieve an AUC of 0.99 for COVID-19, 0.97 for normal, and 0.97 for pneumonia. The model was able to outperform alternative models in terms of overall accuracy, sensitivity, and specificity.
Conclusions: Our proposed automated diagnostic model yielded an accuracy of 94% in the initial screening of COVID-19 patients and an overall accuracy of 92.19% using chest x-ray images.
COVID-19 is a highly contagious infectious disease that has infected millions of people worldwide. Polymerase Chain Reaction (PCR) is the gold standard diagnostic test available for COVID-19 detection. Alternatively, medical imaging techniques, including chest X-ray (CXR), has been instrumental in diagnosis and prognosis of patients with COVID-19. Enabling the CXR with machine learning-based automated diagnosis will be important for rapid diagnosis of the disease by minimizing manual assessment of images by the radiologists. In this work, we developed a deep learning model that utilizes the transfer learning approach using a pre-trained Residual Network model. The Residual Network 50 (ResNet50) is trained from scratch by utilizing the initial architecture and pre-trained weights to provide the classification results. Two types of classification (two-class and three-class) is performed using the developed model. A cascaded approach is adopted for two-class classification where the classification is performed in two phases. The dataset used for training and evaluating the model comprises of 8,254 images in total out of which 1651 images were considered for testing the cascaded model (15 COVID-19) and three-class classification (51 COVID-19). The model was evaluated using accuracy, sensitivity, specificity, and F1-score metrics. Our cascaded model yielded an accuracy of 91.8% for classification of abnormal and normal cases and 97.9% for the classification of pneumonia and COVID-19 images. In the three-class classification, our model reported an accuracy of 92% in classifying normal, pneumonia (bacterial and viral) and COVID-19 cases.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.