File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Deep Mining External Imperfect Data for Chest X-Ray Disease Screening

TitleDeep Mining External Imperfect Data for Chest X-Ray Disease Screening
Authors
Issue Date2020
Citation
IEEE Transactions on Medical Imaging, 2020, v. 39, n. 11, p. 3583-3594 How to Cite?
AbstractDeep learning approaches have demonstrated remarkable progress in automatic Chest X-ray analysis. The data-driven feature of deep models requires training data to cover a large distribution. Therefore, it is substantial to integrate knowledge from multiple datasets, especially for medical images. However, learning a disease classification model with extra Chest X-ray (CXR) data is yet challenging. Recent researches have demonstrated that performance bottleneck exists in joint training on different CXR datasets, and few made efforts to address the obstacle. In this paper, we argue that incorporating an external CXR dataset leads to imperfect training data, which raises the challenges. Specifically, the imperfect data is in two folds: domain discrepancy, as the image appearances vary across datasets; and label discrepancy, as different datasets are partially labeled. To this end, we formulate the multi-label thoracic disease classification problem as weighted independent binary tasks according to the categories. For common categories shared across domains, we adopt task-specific adversarial training to alleviate the feature differences. For categories existing in a single dataset, we present uncertainty-aware temporal ensembling of model predictions to mine the information from the missing labels further. In this way, our framework simultaneously models and tackles the domain and label discrepancies, enabling superior knowledge mining ability. We conduct extensive experiments on three datasets with more than 360,000 Chest X-ray images. Our method outperforms other competing models and sets state-of-the-art performance on the official NIH test set with 0.8349 AUC, demonstrating its effectiveness of utilizing the external dataset to improve the internal classification.
Persistent Identifierhttp://hdl.handle.net/10722/299477
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorLuo, Luyang-
dc.contributor.authorYu, Lequan-
dc.contributor.authorChen, Hao-
dc.contributor.authorLiu, Quande-
dc.contributor.authorWang, Xi-
dc.contributor.authorXu, Jiaqi-
dc.contributor.authorHeng, Pheng Ann-
dc.date.accessioned2021-05-21T03:34:29Z-
dc.date.available2021-05-21T03:34:29Z-
dc.date.issued2020-
dc.identifier.citationIEEE Transactions on Medical Imaging, 2020, v. 39, n. 11, p. 3583-3594-
dc.identifier.urihttp://hdl.handle.net/10722/299477-
dc.description.abstractDeep learning approaches have demonstrated remarkable progress in automatic Chest X-ray analysis. The data-driven feature of deep models requires training data to cover a large distribution. Therefore, it is substantial to integrate knowledge from multiple datasets, especially for medical images. However, learning a disease classification model with extra Chest X-ray (CXR) data is yet challenging. Recent researches have demonstrated that performance bottleneck exists in joint training on different CXR datasets, and few made efforts to address the obstacle. In this paper, we argue that incorporating an external CXR dataset leads to imperfect training data, which raises the challenges. Specifically, the imperfect data is in two folds: domain discrepancy, as the image appearances vary across datasets; and label discrepancy, as different datasets are partially labeled. To this end, we formulate the multi-label thoracic disease classification problem as weighted independent binary tasks according to the categories. For common categories shared across domains, we adopt task-specific adversarial training to alleviate the feature differences. For categories existing in a single dataset, we present uncertainty-aware temporal ensembling of model predictions to mine the information from the missing labels further. In this way, our framework simultaneously models and tackles the domain and label discrepancies, enabling superior knowledge mining ability. We conduct extensive experiments on three datasets with more than 360,000 Chest X-ray images. Our method outperforms other competing models and sets state-of-the-art performance on the official NIH test set with 0.8349 AUC, demonstrating its effectiveness of utilizing the external dataset to improve the internal classification.-
dc.languageeng-
dc.relation.ispartofIEEE Transactions on Medical Imaging-
dc.titleDeep Mining External Imperfect Data for Chest X-Ray Disease Screening-
dc.typeArticle-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1109/TMI.2020.3000949-
dc.identifier.pmid32746106-
dc.identifier.scopuseid_2-s2.0-85094933040-
dc.identifier.volume39-
dc.identifier.issue11-
dc.identifier.spage3583-
dc.identifier.epage3594-
dc.identifier.eissn1558-254X-
dc.identifier.isiWOS:000586352000029-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats