Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations
Deep learning (DL) has drawn tremendous attention for object localization and recognition in both natural and medical images. U-Net segmentation models have demonstrated superior performance compared to conventional hand-crafted feature-based methods. Medical image modality-specific DL models are be...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-03-01
|
| Series: | Diagnostics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-4418/11/4/616 |
| _version_ | 1827696411759607808 |
|---|---|
| author | Sivaramakrishnan Rajaraman Les R. Folio Jane Dimperio Philip O. Alderson Sameer K. Antani |
| author_facet | Sivaramakrishnan Rajaraman Les R. Folio Jane Dimperio Philip O. Alderson Sameer K. Antani |
| author_sort | Sivaramakrishnan Rajaraman |
| collection | DOAJ |
| description | Deep learning (DL) has drawn tremendous attention for object localization and recognition in both natural and medical images. U-Net segmentation models have demonstrated superior performance compared to conventional hand-crafted feature-based methods. Medical image modality-specific DL models are better at transferring domain knowledge to a relevant target task than those pretrained on stock photography images. This character helps improve model adaptation, generalization, and class-specific region of interest (ROI) localization. In this study, we train chest X-ray (CXR) modality-specific U-Nets and other state-of-the-art U-Net models for semantic segmentation of tuberculosis (TB)-consistent findings. Automated segmentation of such manifestations could help radiologists reduce errors and supplement decision-making while improving patient care and productivity. Our approach uses the publicly available TBX11K CXR dataset with weak TB annotations, typically provided as bounding boxes, to train a set of U-Net models. Next, we improve the results by augmenting the training data with weak localization, postprocessed into an ROI mask, from a DL classifier trained to classify CXRs as showing normal lungs or suspected TB manifestations. Test data are individually derived from the TBX11K CXR training distribution and other cross-institutional collections, including the Shenzhen TB and Montgomery TB CXR datasets. We observe that our augmented training strategy helped the CXR modality-specific U-Net models achieve superior performance with test data derived from the TBX11K CXR training distribution and cross-institutional collections (<i>p</i> < 0.05). We believe that this is the first study to i) use CXR modality-specific U-Nets for semantic segmentation of TB-consistent ROIs and ii) evaluate the segmentation performance while augmenting the training data with weak TB-consistent localizations. |
| first_indexed | 2024-03-10T12:46:22Z |
| format | Article |
| id | doaj.art-45b8da0e64d64853bc5ca9a3a4cbad73 |
| institution | Directory Open Access Journal |
| issn | 2075-4418 |
| language | English |
| last_indexed | 2024-03-10T12:46:22Z |
| publishDate | 2021-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Diagnostics |
| spelling | doaj.art-45b8da0e64d64853bc5ca9a3a4cbad732023-11-21T13:24:24ZengMDPI AGDiagnostics2075-44182021-03-0111461610.3390/diagnostics11040616Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak LocalizationsSivaramakrishnan Rajaraman0Les R. Folio1Jane Dimperio2Philip O. Alderson3Sameer K. Antani4National Library of Medicine, National Institutes of Health, Bethesda, MD 20894, USARadiology and Imaging Sciences, Clinical Center, National Institutes of Health, Bethesda, MD 20894, USARadiology and Imaging Sciences, Clinical Center, National Institutes of Health, Bethesda, MD 20894, USASchool of Medicine, Saint Louis University, St. Louis, MO 63104, USANational Library of Medicine, National Institutes of Health, Bethesda, MD 20894, USADeep learning (DL) has drawn tremendous attention for object localization and recognition in both natural and medical images. U-Net segmentation models have demonstrated superior performance compared to conventional hand-crafted feature-based methods. Medical image modality-specific DL models are better at transferring domain knowledge to a relevant target task than those pretrained on stock photography images. This character helps improve model adaptation, generalization, and class-specific region of interest (ROI) localization. In this study, we train chest X-ray (CXR) modality-specific U-Nets and other state-of-the-art U-Net models for semantic segmentation of tuberculosis (TB)-consistent findings. Automated segmentation of such manifestations could help radiologists reduce errors and supplement decision-making while improving patient care and productivity. Our approach uses the publicly available TBX11K CXR dataset with weak TB annotations, typically provided as bounding boxes, to train a set of U-Net models. Next, we improve the results by augmenting the training data with weak localization, postprocessed into an ROI mask, from a DL classifier trained to classify CXRs as showing normal lungs or suspected TB manifestations. Test data are individually derived from the TBX11K CXR training distribution and other cross-institutional collections, including the Shenzhen TB and Montgomery TB CXR datasets. We observe that our augmented training strategy helped the CXR modality-specific U-Net models achieve superior performance with test data derived from the TBX11K CXR training distribution and cross-institutional collections (<i>p</i> < 0.05). We believe that this is the first study to i) use CXR modality-specific U-Nets for semantic segmentation of TB-consistent ROIs and ii) evaluate the segmentation performance while augmenting the training data with weak TB-consistent localizations.https://www.mdpi.com/2075-4418/11/4/616deep learningtuberculosisconvolutional neural networkssegmentationmodality-specific knowledge transferU-Net |
| spellingShingle | Sivaramakrishnan Rajaraman Les R. Folio Jane Dimperio Philip O. Alderson Sameer K. Antani Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations Diagnostics deep learning tuberculosis convolutional neural networks segmentation modality-specific knowledge transfer U-Net |
| title | Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations |
| title_full | Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations |
| title_fullStr | Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations |
| title_full_unstemmed | Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations |
| title_short | Improved Semantic Segmentation of Tuberculosis—Consistent Findings in Chest X-rays Using Augmented Training of Modality-Specific U-Net Models with Weak Localizations |
| title_sort | improved semantic segmentation of tuberculosis consistent findings in chest x rays using augmented training of modality specific u net models with weak localizations |
| topic | deep learning tuberculosis convolutional neural networks segmentation modality-specific knowledge transfer U-Net |
| url | https://www.mdpi.com/2075-4418/11/4/616 |
| work_keys_str_mv | AT sivaramakrishnanrajaraman improvedsemanticsegmentationoftuberculosisconsistentfindingsinchestxraysusingaugmentedtrainingofmodalityspecificunetmodelswithweaklocalizations AT lesrfolio improvedsemanticsegmentationoftuberculosisconsistentfindingsinchestxraysusingaugmentedtrainingofmodalityspecificunetmodelswithweaklocalizations AT janedimperio improvedsemanticsegmentationoftuberculosisconsistentfindingsinchestxraysusingaugmentedtrainingofmodalityspecificunetmodelswithweaklocalizations AT philipoalderson improvedsemanticsegmentationoftuberculosisconsistentfindingsinchestxraysusingaugmentedtrainingofmodalityspecificunetmodelswithweaklocalizations AT sameerkantani improvedsemanticsegmentationoftuberculosisconsistentfindingsinchestxraysusingaugmentedtrainingofmodalityspecificunetmodelswithweaklocalizations |