{
  "id": 207675,
  "title": "Papers on X-Rays and CXR",
  "url": "/competitions/vinbigdata-chest-xray-abnormalities-detection/discussion/207675",
  "author_name": "Charlie Craine",
  "post_date": "2020-12-30T19:54:21.083000",
  "votes": 28,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Hey everyone!</p>\n<p>I wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!</p>\n<p>Research Papers:</p>\n<ul>\n<li><p><a href=\"https://arxiv.org/abs/2012.12712\" target=\"_blank\">Chest x-ray automated triage: a semiologic approach designed for clinical implementation, exploiting different types of labels through a combination of four Deep Learning architectures</a> - RESULTS: The external and local test sets had 4376 and 1064 images, respectively, for which the model showed an area under the Receiver Operating Characteristics curve of 0.75 (95%CI: 0.74-0.76) and 0.87 (95%CI: 0.86-0.89) in the detection of abnormal chest x-rays. For the local population, a sensitivity of 86% (95%CI: 84-90), and a specificity of 88% (95%CI: 86-90) were obtained, with no significant differences between demographic subgroups. We present examples of heatmaps to show the accomplished level of interpretability, examining true and false positives.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.14204\" target=\"_blank\">Screening COVID-19 Based on CT/CXR Images &amp; Building a Publicly Available CT-scan Dataset of COVID-19</a> - his study firstly builds a large-size publicly available CT-scan dataset, consisting of more than 13k CT-images of more than 1000 individuals, in which 8k images are taken from 500 patients infected with COVID-19. Secondly, we propose a deep learning model for screening COVID-19 using our proposed CT dataset and report the baseline results. Finally, we extend the proposed CT model for screening COVID-19 from CXR images using a transfer learning approach. The experimental results show that the proposed CT and CXR methods achieve the AUC scores of 0.886 and 0.984 respectively.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.13605\" target=\"_blank\">COVIDX: Computer-aided diagnosis of Covid-19 and its severity prediction with raw digital chest X-ray images</a> - In this study, we present an automatic COVID-19 diagnostic and severity prediction (COVIDX) system that uses deep feature maps from CXR images to diagnose COVID-19 and its severity prediction.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.11911\" target=\"_blank\">A Hybrid VDV Model for Automatic Diagnosis of Pneumothorax using Class-Imbalanced Chest X-rays Dataset</a> - Our proposed framework is tested on SIIM ACR Pneumothorax dataset and Random Sample of NIH Chest X-ray dataset (RS-NIH). For the first dataset, 85.17% Recall with 86.0% Area under the Receiver Operating Characteristic curve (AUC) is attained. For the second dataset, 90.9% Recall with 95.0% AUC is achieved with random split of data while 85.45% recall with 77.06% AUC is obtained with patient-wise split of data.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.10564\" target=\"_blank\">Computer-aided abnormality detection in chest radiographs in a clinical setting via domain-adaptation</a> - In the machine learning community, the challenges posed by the heterogeneity in the data generation source is known as domain shift, which is a mode shift in the generative model. In this work, we introduce a domain-shift detection and removal method to overcome this problem. Our experimental results show the proposed method's effectiveness in deploying a pre-trained DL model for abnormality detection in chest radiographs in a clinical setting.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.07332\" target=\"_blank\">Combining Similarity and Adversarial Learning to Generate Visual Explanation: Application to Medical Image Classification</a> - Finally, we show that random geometric augmentations applied to the original image play a regularization role that improves several previously proposed explanation methods. We validate our approach on a large chest X-ray database.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.06346\" target=\"_blank\">Distant Domain Transfer Learning for Medical Imaging</a> - The main contributions of this study: 1) the proposed method benefits from unlabeled data collected from distant domains which can be easily accessed, 2) it can effectively handle the distribution shift between the training data and the testing data, 3) it has achieved 96\\% classification accuracy, which is 13\\% higher classification accuracy than \"non-transfer\" algorithms, and 8\\% higher than existing transfer and distant transfer algorithms.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.05064\" target=\"_blank\">Secure Medical Image Analysis with CrypTFlow</a> - We empirically demonstrate the power of our system by showing the secure inference of real-world neural networks such as DENSENET121 for detection of lung diseases from chest X-ray images and 3D-UNet for segmentation in radiotherapy planning using CT images. In particular, this paper provides the first evaluation of secure segmentation of 3D images, a task that requires much more powerful models than classification and is the largest secure inference task run till date.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.12734\" target=\"_blank\">Interpreting Chest X-rays via CNNs that Exploit Hierarchical Disease Dependencies and Uncertainty Labels</a> - Our deep net-works were trained on over 200,000 CXRs of the recently released CheXpert dataset (Irvinandal., 2019) and the final model, which was an ensemble of the best performing networks,achieved a mean area under the curve (AUC) of 0.940 in predicting 5 selected pathologiesfrom the validation set. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.05922\" target=\"_blank\">Continual Learning for Domain Adaptation in Chest X-ray Classification</a> - Using the ChestX-ray14 and the MIMIC-CXR datasets, we demonstrate empirically that these methods provide promising options to improve the performance of Deep Learning models on a target domain and to mitigate effectively catastrophic forgetting for the source domain. To this end, the best overall performance was obtained using JT, while for LWF competitive results could be achieved - even without accessing data from the source domain.</p></li>\n</ul>",
  "messages": [
    {
      "id": 1132938,
      "postDate": "2020-12-30T19:54:21.083Z",
      "content": "<p>Hey everyone!</p>\n<p>I wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!</p>\n<p>Research Papers:</p>\n<ul>\n<li><p><a href=\"https://arxiv.org/abs/2012.12712\" target=\"_blank\">Chest x-ray automated triage: a semiologic approach designed for clinical implementation, exploiting different types of labels through a combination of four Deep Learning architectures</a> - RESULTS: The external and local test sets had 4376 and 1064 images, respectively, for which the model showed an area under the Receiver Operating Characteristics curve of 0.75 (95%CI: 0.74-0.76) and 0.87 (95%CI: 0.86-0.89) in the detection of abnormal chest x-rays. For the local population, a sensitivity of 86% (95%CI: 84-90), and a specificity of 88% (95%CI: 86-90) were obtained, with no significant differences between demographic subgroups. We present examples of heatmaps to show the accomplished level of interpretability, examining true and false positives.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.14204\" target=\"_blank\">Screening COVID-19 Based on CT/CXR Images &amp; Building a Publicly Available CT-scan Dataset of COVID-19</a> - his study firstly builds a large-size publicly available CT-scan dataset, consisting of more than 13k CT-images of more than 1000 individuals, in which 8k images are taken from 500 patients infected with COVID-19. Secondly, we propose a deep learning model for screening COVID-19 using our proposed CT dataset and report the baseline results. Finally, we extend the proposed CT model for screening COVID-19 from CXR images using a transfer learning approach. The experimental results show that the proposed CT and CXR methods achieve the AUC scores of 0.886 and 0.984 respectively.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.13605\" target=\"_blank\">COVIDX: Computer-aided diagnosis of Covid-19 and its severity prediction with raw digital chest X-ray images</a> - In this study, we present an automatic COVID-19 diagnostic and severity prediction (COVIDX) system that uses deep feature maps from CXR images to diagnose COVID-19 and its severity prediction.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.11911\" target=\"_blank\">A Hybrid VDV Model for Automatic Diagnosis of Pneumothorax using Class-Imbalanced Chest X-rays Dataset</a> - Our proposed framework is tested on SIIM ACR Pneumothorax dataset and Random Sample of NIH Chest X-ray dataset (RS-NIH). For the first dataset, 85.17% Recall with 86.0% Area under the Receiver Operating Characteristic curve (AUC) is attained. For the second dataset, 90.9% Recall with 95.0% AUC is achieved with random split of data while 85.45% recall with 77.06% AUC is obtained with patient-wise split of data.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.10564\" target=\"_blank\">Computer-aided abnormality detection in chest radiographs in a clinical setting via domain-adaptation</a> - In the machine learning community, the challenges posed by the heterogeneity in the data generation source is known as domain shift, which is a mode shift in the generative model. In this work, we introduce a domain-shift detection and removal method to overcome this problem. Our experimental results show the proposed method's effectiveness in deploying a pre-trained DL model for abnormality detection in chest radiographs in a clinical setting.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.07332\" target=\"_blank\">Combining Similarity and Adversarial Learning to Generate Visual Explanation: Application to Medical Image Classification</a> - Finally, we show that random geometric augmentations applied to the original image play a regularization role that improves several previously proposed explanation methods. We validate our approach on a large chest X-ray database.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.06346\" target=\"_blank\">Distant Domain Transfer Learning for Medical Imaging</a> - The main contributions of this study: 1) the proposed method benefits from unlabeled data collected from distant domains which can be easily accessed, 2) it can effectively handle the distribution shift between the training data and the testing data, 3) it has achieved 96\\% classification accuracy, which is 13\\% higher classification accuracy than \"non-transfer\" algorithms, and 8\\% higher than existing transfer and distant transfer algorithms.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2012.05064\" target=\"_blank\">Secure Medical Image Analysis with CrypTFlow</a> - We empirically demonstrate the power of our system by showing the secure inference of real-world neural networks such as DENSENET121 for detection of lung diseases from chest X-ray images and 3D-UNet for segmentation in radiotherapy planning using CT images. In particular, this paper provides the first evaluation of secure segmentation of 3D images, a task that requires much more powerful models than classification and is the largest secure inference task run till date.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.12734\" target=\"_blank\">Interpreting Chest X-rays via CNNs that Exploit Hierarchical Disease Dependencies and Uncertainty Labels</a> - Our deep net-works were trained on over 200,000 CXRs of the recently released CheXpert dataset (Irvinandal., 2019) and the final model, which was an ensemble of the best performing networks,achieved a mean area under the curve (AUC) of 0.940 in predicting 5 selected pathologiesfrom the validation set. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.05922\" target=\"_blank\">Continual Learning for Domain Adaptation in Chest X-ray Classification</a> - Using the ChestX-ray14 and the MIMIC-CXR datasets, we demonstrate empirically that these methods provide promising options to improve the performance of Deep Learning models on a target domain and to mitigate effectively catastrophic forgetting for the source domain. To this end, the best overall performance was obtained using JT, while for LWF competitive results could be achieved - even without accessing data from the source domain.</p></li>\n</ul>",
      "rawMarkdown": "Hey everyone!\n\nI wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!\n\nResearch Papers:\n\n* [Chest x-ray automated triage: a semiologic approach designed for clinical implementation, exploiting different types of labels through a combination of four Deep Learning architectures](https://arxiv.org/abs/2012.12712) - RESULTS: The external and local test sets had 4376 and 1064 images, respectively, for which the model showed an area under the Receiver Operating Characteristics curve of 0.75 (95%CI: 0.74-0.76) and 0.87 (95%CI: 0.86-0.89) in the detection of abnormal chest x-rays. For the local population, a sensitivity of 86% (95%CI: 84-90), and a specificity of 88% (95%CI: 86-90) were obtained, with no significant differences between demographic subgroups. We present examples of heatmaps to show the accomplished level of interpretability, examining true and false positives.\n\n* [Screening COVID-19 Based on CT/CXR Images & Building a Publicly Available CT-scan Dataset of COVID-19](https://arxiv.org/abs/2012.14204) - his study firstly builds a large-size publicly available CT-scan dataset, consisting of more than 13k CT-images of more than 1000 individuals, in which 8k images are taken from 500 patients infected with COVID-19. Secondly, we propose a deep learning model for screening COVID-19 using our proposed CT dataset and report the baseline results. Finally, we extend the proposed CT model for screening COVID-19 from CXR images using a transfer learning approach. The experimental results show that the proposed CT and CXR methods achieve the AUC scores of 0.886 and 0.984 respectively.\n\n* [COVIDX: Computer-aided diagnosis of Covid-19 and its severity prediction with raw digital chest X-ray images](https://arxiv.org/abs/2012.13605) - In this study, we present an automatic COVID-19 diagnostic and severity prediction (COVIDX) system that uses deep feature maps from CXR images to diagnose COVID-19 and its severity prediction.\n\n* [A Hybrid VDV Model for Automatic Diagnosis of Pneumothorax using Class-Imbalanced Chest X-rays Dataset](https://arxiv.org/abs/2012.11911) - Our proposed framework is tested on SIIM ACR Pneumothorax dataset and Random Sample of NIH Chest X-ray dataset (RS-NIH). For the first dataset, 85.17% Recall with 86.0% Area under the Receiver Operating Characteristic curve (AUC) is attained. For the second dataset, 90.9% Recall with 95.0% AUC is achieved with random split of data while 85.45% recall with 77.06% AUC is obtained with patient-wise split of data.\n\n* [Computer-aided abnormality detection in chest radiographs in a clinical setting via domain-adaptation](https://arxiv.org/abs/2012.10564) - In the machine learning community, the challenges posed by the heterogeneity in the data generation source is known as domain shift, which is a mode shift in the generative model. In this work, we introduce a domain-shift detection and removal method to overcome this problem. Our experimental results show the proposed method's effectiveness in deploying a pre-trained DL model for abnormality detection in chest radiographs in a clinical setting.\n\n* [Combining Similarity and Adversarial Learning to Generate Visual Explanation: Application to Medical Image Classification](https://arxiv.org/abs/2012.07332) - Finally, we show that random geometric augmentations applied to the original image play a regularization role that improves several previously proposed explanation methods. We validate our approach on a large chest X-ray database.\n\n* [Distant Domain Transfer Learning for Medical Imaging](https://arxiv.org/abs/2012.06346) - The main contributions of this study: 1) the proposed method benefits from unlabeled data collected from distant domains which can be easily accessed, 2) it can effectively handle the distribution shift between the training data and the testing data, 3) it has achieved 96\\% classification accuracy, which is 13\\% higher classification accuracy than \"non-transfer\" algorithms, and 8\\% higher than existing transfer and distant transfer algorithms.\n\n* [Secure Medical Image Analysis with CrypTFlow](https://arxiv.org/abs/2012.05064) - We empirically demonstrate the power of our system by showing the secure inference of real-world neural networks such as DENSENET121 for detection of lung diseases from chest X-ray images and 3D-UNet for segmentation in radiotherapy planning using CT images. In particular, this paper provides the first evaluation of secure segmentation of 3D images, a task that requires much more powerful models than classification and is the largest secure inference task run till date.\n\n* [Interpreting Chest X-rays via CNNs that Exploit Hierarchical Disease Dependencies and Uncertainty Labels](https://arxiv.org/abs/2005.12734) - Our deep net-works were trained on over 200,000 CXRs of the recently released CheXpert dataset (Irvinandal., 2019) and the final model, which was an ensemble of the best performing networks,achieved a mean area under the curve (AUC) of 0.940 in predicting 5 selected pathologiesfrom the validation set. \n\n* [Continual Learning for Domain Adaptation in Chest X-ray Classification](https://arxiv.org/abs/2001.05922) - Using the ChestX-ray14 and the MIMIC-CXR datasets, we demonstrate empirically that these methods provide promising options to improve the performance of Deep Learning models on a target domain and to mitigate effectively catastrophic forgetting for the source domain. To this end, the best overall performance was obtained using JT, while for LWF competitive results could be achieved - even without accessing data from the source domain.\n",
      "votes": 28
    },
    {
      "id": 1132943,
      "postDate": "2020-12-30T20:02:33.430Z",
      "content": "<p>Add the organizer's papers as well, as it explains the creation methodologies of the dataset.</p>\n<p>“VinDr-CXR: An open dataset of chest X-rays with radiologist's annotations”.: <a href=\"https://storage.googleapis.com/kaggle-media/competitions/VinBigData/VinDr_CXR_data_paper.pdf\" target=\"_blank\">https://storage.googleapis.com/kaggle-media/competitions/VinBigData/VinDr_CXR_data_paper.pdf</a></p>\n<p>Though it's there on the overview page, yet having all the papers in a single thread will be helpful I believe.</p>",
      "rawMarkdown": "Add the organizer's papers as well, as it explains the creation methodologies of the dataset.\n\n“VinDr-CXR: An open dataset of chest X-rays with radiologist's annotations”.: https://storage.googleapis.com/kaggle-media/competitions/VinBigData/VinDr_CXR_data_paper.pdf\n\nThough it's there on the overview page, yet having all the papers in a single thread will be helpful I believe.",
      "votes": 4
    },
    {
      "id": 1149715,
      "postDate": "2021-01-12T04:49:37.617Z",
      "content": "<p>I found yet another useful dataset. Hope it helps you too!</p>\n<p><strong>TBX11k Dataset (Tuberculosis Classification and detection)</strong><br>\nThis dataset contains about 11K chest x-ray images with train and valid split. The class labels in TB x-rays are of (TB, Non-TB, Sick but not TB). It also has the TB area detection annotations.<br>\nResearch paper: <a href=\"https://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Rethinking_Computer-Aided_Tuberculosis_Diagnosis_CVPR_2020_paper.html\" target=\"_blank\">https://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Rethinking_Computer-Aided_Tuberculosis_Diagnosis_CVPR_2020_paper.html</a><br>\nDataset competition: <a href=\"https://competitions.codalab.org/competitions/25848\" target=\"_blank\">https://competitions.codalab.org/competitions/25848</a></p>",
      "rawMarkdown": "I found yet another useful dataset. Hope it helps you too!\n\n**TBX11k Dataset (Tuberculosis Classification and detection)**\nThis dataset contains about 11K chest x-ray images with train and valid split. The class labels in TB x-rays are of (TB, Non-TB, Sick but not TB). It also has the TB area detection annotations.\nResearch paper: https://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Rethinking_Computer-Aided_Tuberculosis_Diagnosis_CVPR_2020_paper.html\nDataset competition: https://competitions.codalab.org/competitions/25848",
      "votes": 2
    },
    {
      "id": 1139811,
      "postDate": "2021-01-05T16:27:24.833Z",
      "content": "<p><strong>ChestNet: A Deep Neural Network for Classification of\nThoracic Diseases on Chest Radiography</strong></p>\n<p>Maybe this can also be included.</p>",
      "rawMarkdown": "**ChestNet: A Deep Neural Network for Classification of\nThoracic Diseases on Chest Radiography**\n\nMaybe this can also be included.",
      "votes": 2
    },
    {
      "id": 1250015,
      "postDate": "2021-03-23T17:54:23.997Z",
      "content": "<p>This paper '<a href=\"https://arxiv.org/abs/2003.00827\" target=\"_blank\">CheXclusion: Fairness gaps in deep chest X-ray classifiers</a>' has the classification performance of MIMIC-CXR  (average AUC across 14 labels: 0.834±0.001 ), CheXpert (average AUC across 14 labels: 0.805±0.001 ), NIH (ChestX-ray14) dataset  (average AUC across  14 labels: 0.840±0.001), and aggregation of MIMIC-CXR, CheXpert, and NIH on shared labels, called ALL dataset with 707K images (average AUC across 8 labels:0.859±0.001). <br>\nThe paper and code are avialble:<br>\nTitle: CheXclusion: Fairness gaps in deep chest X-ray classifiers<br>\nPaper: <a href=\"https://arxiv.org/pdf/2003.00827.pdf\" target=\"_blank\">https://arxiv.org/pdf/2003.00827.pdf</a><br>\nCode: <a href=\"https://github.com/LalehSeyyed/CheXclusion\" target=\"_blank\">https://github.com/LalehSeyyed/CheXclusion</a></p>",
      "rawMarkdown": "This paper '[CheXclusion: Fairness gaps in deep chest X-ray classifiers](https://arxiv.org/abs/2003.00827)' has the classification performance of MIMIC-CXR  (average AUC across 14 labels: 0.834±0.001 ), CheXpert (average AUC across 14 labels: 0.805±0.001 ), NIH (ChestX-ray14) dataset  (average AUC across  14 labels: 0.840±0.001), and aggregation of MIMIC-CXR, CheXpert, and NIH on shared labels, called ALL dataset with 707K images (average AUC across 8 labels:0.859±0.001). \nThe paper and code are avialble:\nTitle: CheXclusion: Fairness gaps in deep chest X-ray classifiers\nPaper: https://arxiv.org/pdf/2003.00827.pdf\nCode: https://github.com/LalehSeyyed/CheXclusion"
    },
    {
      "id": 1255601,
      "postDate": "2021-03-29T00:36:41.387Z",
      "content": "<p>Thanks for sharing! :)</p>",
      "rawMarkdown": "Thanks for sharing! :)"
    },
    {
      "id": 1215498,
      "postDate": "2021-02-23T17:01:12.717Z",
      "content": "<p>Thanks for sharing this, very useful!</p>",
      "rawMarkdown": "Thanks for sharing this, very useful!"
    }
  ],
  "comments": [
    {
      "id": 1132943,
      "author_name": "Ultron",
      "author_url": "",
      "post_date": "2020-12-30T20:02:33.430000",
      "content": "<p>Add the organizer's papers as well, as it explains the creation methodologies of the dataset.</p>\n<p>“VinDr-CXR: An open dataset of chest X-rays with radiologist's annotations”.: <a href=\"https://storage.googleapis.com/kaggle-media/competitions/VinBigData/VinDr_CXR_data_paper.pdf\" target=\"_blank\">https://storage.googleapis.com/kaggle-media/competitions/VinBigData/VinDr_CXR_data_paper.pdf</a></p>\n<p>Though it's there on the overview page, yet having all the papers in a single thread will be helpful I believe.</p>",
      "votes": 4,
      "replies": []
    },
    {
      "id": 1149715,
      "author_name": "HARINI NARASIMHAN",
      "author_url": "",
      "post_date": "2021-01-12T04:49:37.617000",
      "content": "<p>I found yet another useful dataset. Hope it helps you too!</p>\n<p><strong>TBX11k Dataset (Tuberculosis Classification and detection)</strong><br>\nThis dataset contains about 11K chest x-ray images with train and valid split. The class labels in TB x-rays are of (TB, Non-TB, Sick but not TB). It also has the TB area detection annotations.<br>\nResearch paper: <a href=\"https://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Rethinking_Computer-Aided_Tuberculosis_Diagnosis_CVPR_2020_paper.html\" target=\"_blank\">https://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Rethinking_Computer-Aided_Tuberculosis_Diagnosis_CVPR_2020_paper.html</a><br>\nDataset competition: <a href=\"https://competitions.codalab.org/competitions/25848\" target=\"_blank\">https://competitions.codalab.org/competitions/25848</a></p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 1139811,
      "author_name": "Gursewak Dhiman",
      "author_url": "",
      "post_date": "2021-01-05T16:27:24.833000",
      "content": "<p><strong>ChestNet: A Deep Neural Network for Classification of\nThoracic Diseases on Chest Radiography</strong></p>\n<p>Maybe this can also be included.</p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 1250015,
      "author_name": "Laleh",
      "author_url": "",
      "post_date": "2021-03-23T17:54:23.997000",
      "content": "<p>This paper '<a href=\"https://arxiv.org/abs/2003.00827\" target=\"_blank\">CheXclusion: Fairness gaps in deep chest X-ray classifiers</a>' has the classification performance of MIMIC-CXR  (average AUC across 14 labels: 0.834±0.001 ), CheXpert (average AUC across 14 labels: 0.805±0.001 ), NIH (ChestX-ray14) dataset  (average AUC across  14 labels: 0.840±0.001), and aggregation of MIMIC-CXR, CheXpert, and NIH on shared labels, called ALL dataset with 707K images (average AUC across 8 labels:0.859±0.001). <br>\nThe paper and code are avialble:<br>\nTitle: CheXclusion: Fairness gaps in deep chest X-ray classifiers<br>\nPaper: <a href=\"https://arxiv.org/pdf/2003.00827.pdf\" target=\"_blank\">https://arxiv.org/pdf/2003.00827.pdf</a><br>\nCode: <a href=\"https://github.com/LalehSeyyed/CheXclusion\" target=\"_blank\">https://github.com/LalehSeyyed/CheXclusion</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1255601,
      "author_name": "antoreepjana",
      "author_url": "",
      "post_date": "2021-03-29T00:36:41.387000",
      "content": "<p>Thanks for sharing! :)</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1215498,
      "author_name": "Old Monk",
      "author_url": "",
      "post_date": "2021-02-23T17:01:12.717000",
      "content": "<p>Thanks for sharing this, very useful!</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1132938": "Hey everyone!\n\nI wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!\n\nResearch Papers:\n\n* [Chest x-ray automated triage: a semiologic approach designed for clinical implementation, exploiting different types of labels through a combination of four Deep Learning architectures](https://arxiv.org/abs/2012.12712) - RESULTS: The external and local test sets had 4376 and 1064 images, respectively, for which the model showed an area under the Receiver Operating Characteristics curve of 0.75 (95%CI: 0.74-0.76) and 0.87 (95%CI: 0.86-0.89) in the detection of abnormal chest x-rays. For the local population, a sensitivity of 86% (95%CI: 84-90), and a specificity of 88% (95%CI: 86-90) were obtained, with no significant differences between demographic subgroups. We present examples of heatmaps to show the accomplished level of interpretability, examining true and false positives.\n\n* [Screening COVID-19 Based on CT/CXR Images & Building a Publicly Available CT-scan Dataset of COVID-19](https://arxiv.org/abs/2012.14204) - his study firstly builds a large-size publicly available CT-scan dataset, consisting of more than 13k CT-images of more than 1000 individuals, in which 8k images are taken from 500 patients infected with COVID-19. Secondly, we propose a deep learning model for screening COVID-19 using our proposed CT dataset and report the baseline results. Finally, we extend the proposed CT model for screening COVID-19 from CXR images using a transfer learning approach. The experimental results show that the proposed CT and CXR methods achieve the AUC scores of 0.886 and 0.984 respectively.\n\n* [COVIDX: Computer-aided diagnosis of Covid-19 and its severity prediction with raw digital chest X-ray images](https://arxiv.org/abs/2012.13605) - In this study, we present an automatic COVID-19 diagnostic and severity prediction (COVIDX) system that uses deep feature maps from CXR images to diagnose COVID-19 and its severity prediction.\n\n* [A Hybrid VDV Model for Automatic Diagnosis of Pneumothorax using Class-Imbalanced Chest X-rays Dataset](https://arxiv.org/abs/2012.11911) - Our proposed framework is tested on SIIM ACR Pneumothorax dataset and Random Sample of NIH Chest X-ray dataset (RS-NIH). For the first dataset, 85.17% Recall with 86.0% Area under the Receiver Operating Characteristic curve (AUC) is attained. For the second dataset, 90.9% Recall with 95.0% AUC is achieved with random split of data while 85.45% recall with 77.06% AUC is obtained with patient-wise split of data.\n\n* [Computer-aided abnormality detection in chest radiographs in a clinical setting via domain-adaptation](https://arxiv.org/abs/2012.10564) - In the machine learning community, the challenges posed by the heterogeneity in the data generation source is known as domain shift, which is a mode shift in the generative model. In this work, we introduce a domain-shift detection and removal method to overcome this problem. Our experimental results show the proposed method's effectiveness in deploying a pre-trained DL model for abnormality detection in chest radiographs in a clinical setting.\n\n* [Combining Similarity and Adversarial Learning to Generate Visual Explanation: Application to Medical Image Classification](https://arxiv.org/abs/2012.07332) - Finally, we show that random geometric augmentations applied to the original image play a regularization role that improves several previously proposed explanation methods. We validate our approach on a large chest X-ray database.\n\n* [Distant Domain Transfer Learning for Medical Imaging](https://arxiv.org/abs/2012.06346) - The main contributions of this study: 1) the proposed method benefits from unlabeled data collected from distant domains which can be easily accessed, 2) it can effectively handle the distribution shift between the training data and the testing data, 3) it has achieved 96\\% classification accuracy, which is 13\\% higher classification accuracy than \"non-transfer\" algorithms, and 8\\% higher than existing transfer and distant transfer algorithms.\n\n* [Secure Medical Image Analysis with CrypTFlow](https://arxiv.org/abs/2012.05064) - We empirically demonstrate the power of our system by showing the secure inference of real-world neural networks such as DENSENET121 for detection of lung diseases from chest X-ray images and 3D-UNet for segmentation in radiotherapy planning using CT images. In particular, this paper provides the first evaluation of secure segmentation of 3D images, a task that requires much more powerful models than classification and is the largest secure inference task run till date.\n\n* [Interpreting Chest X-rays via CNNs that Exploit Hierarchical Disease Dependencies and Uncertainty Labels](https://arxiv.org/abs/2005.12734) - Our deep net-works were trained on over 200,000 CXRs of the recently released CheXpert dataset (Irvinandal., 2019) and the final model, which was an ensemble of the best performing networks,achieved a mean area under the curve (AUC) of 0.940 in predicting 5 selected pathologiesfrom the validation set. \n\n* [Continual Learning for Domain Adaptation in Chest X-ray Classification](https://arxiv.org/abs/2001.05922) - Using the ChestX-ray14 and the MIMIC-CXR datasets, we demonstrate empirically that these methods provide promising options to improve the performance of Deep Learning models on a target domain and to mitigate effectively catastrophic forgetting for the source domain. To this end, the best overall performance was obtained using JT, while for LWF competitive results could be achieved - even without accessing data from the source domain.\n",
    "1132943": "Add the organizer's papers as well, as it explains the creation methodologies of the dataset.\n\n“VinDr-CXR: An open dataset of chest X-rays with radiologist's annotations”.: https://storage.googleapis.com/kaggle-media/competitions/VinBigData/VinDr_CXR_data_paper.pdf\n\nThough it's there on the overview page, yet having all the papers in a single thread will be helpful I believe.",
    "1149715": "I found yet another useful dataset. Hope it helps you too!\n\n**TBX11k Dataset (Tuberculosis Classification and detection)**\nThis dataset contains about 11K chest x-ray images with train and valid split. The class labels in TB x-rays are of (TB, Non-TB, Sick but not TB). It also has the TB area detection annotations.\nResearch paper: https://openaccess.thecvf.com/content_CVPR_2020/html/Liu_Rethinking_Computer-Aided_Tuberculosis_Diagnosis_CVPR_2020_paper.html\nDataset competition: https://competitions.codalab.org/competitions/25848",
    "1139811": "**ChestNet: A Deep Neural Network for Classification of\nThoracic Diseases on Chest Radiography**\n\nMaybe this can also be included.",
    "1250015": "This paper '[CheXclusion: Fairness gaps in deep chest X-ray classifiers](https://arxiv.org/abs/2003.00827)' has the classification performance of MIMIC-CXR  (average AUC across 14 labels: 0.834±0.001 ), CheXpert (average AUC across 14 labels: 0.805±0.001 ), NIH (ChestX-ray14) dataset  (average AUC across  14 labels: 0.840±0.001), and aggregation of MIMIC-CXR, CheXpert, and NIH on shared labels, called ALL dataset with 707K images (average AUC across 8 labels:0.859±0.001). \nThe paper and code are avialble:\nTitle: CheXclusion: Fairness gaps in deep chest X-ray classifiers\nPaper: https://arxiv.org/pdf/2003.00827.pdf\nCode: https://github.com/LalehSeyyed/CheXclusion",
    "1255601": "Thanks for sharing! :)",
    "1215498": "Thanks for sharing this, very useful!"
  }
}