A universal lesion detection method based on partially supervised learning
Partially supervised learning (PSL) is urgently necessary to explore to construct an efficient universal lesion detection (ULD) segmentation model. An annotated dataset is crucial but hard to acquire because of too many Computed tomography (CT) images and the lack of professionals in computer-aided...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2023-08-01
|
| Series: | Frontiers in Pharmacology |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fphar.2023.1084155/full |
| _version_ | 1797767487295586304 |
|---|---|
| author | Xun Wang Xun Wang Xin Shi Xiangyu Meng Zhiyuan Zhang Chaogang Zhang |
| author_facet | Xun Wang Xun Wang Xin Shi Xiangyu Meng Zhiyuan Zhang Chaogang Zhang |
| author_sort | Xun Wang |
| collection | DOAJ |
| description | Partially supervised learning (PSL) is urgently necessary to explore to construct an efficient universal lesion detection (ULD) segmentation model. An annotated dataset is crucial but hard to acquire because of too many Computed tomography (CT) images and the lack of professionals in computer-aided detection/diagnosis (CADe/CADx). To address this problem, we propose a novel loss function to reduce the proportion of negative anchors which is extremely likely to classify the lesion area (positive samples) as a negative bounding box, further leading to an unexpected performance. Before calculating loss, we generate a mask to intentionally choose fewer negative anchors which will backward wrongful loss to the network. During the process of loss calculation, we set a parameter to reduce the proportion of negative samples, and it significantly reduces the adverse effect of misclassification on the model. Our experiments are implemented in a 3D framework by feeding a partially annotated dataset named DeepLesion, a large-scale public dataset for universal lesion detection from CT. We implement a lot of experiments to choose the most suitable parameter, and the result shows that the proposed method has greatly improved the performance of a ULD detector. Our code can be obtained at https://github.com/PLuld0/PLuldl. |
| first_indexed | 2024-03-12T20:40:30Z |
| format | Article |
| id | doaj.art-c4b535c16be5480aa4c9772ee5cacdd3 |
| institution | Directory Open Access Journal |
| issn | 1663-9812 |
| language | English |
| last_indexed | 2024-03-12T20:40:30Z |
| publishDate | 2023-08-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Pharmacology |
| spelling | doaj.art-c4b535c16be5480aa4c9772ee5cacdd32023-08-01T11:01:31ZengFrontiers Media S.A.Frontiers in Pharmacology1663-98122023-08-011410.3389/fphar.2023.10841551084155A universal lesion detection method based on partially supervised learningXun Wang0Xun Wang1Xin Shi2Xiangyu Meng3Zhiyuan Zhang4Chaogang Zhang5Department of Computer Science and Technology, China University of Petroleum, Qingdao, Shandong, ChinaHigh Performance Computer Research Center, University of Chinese Academy of Sciences, Beijing, ChinaDepartment of Computer Science and Technology, China University of Petroleum, Qingdao, Shandong, ChinaDepartment of Computer Science and Technology, China University of Petroleum, Qingdao, Shandong, ChinaDepartment of Computer Science and Technology, China University of Petroleum, Qingdao, Shandong, ChinaDepartment of Computer Science and Technology, China University of Petroleum, Qingdao, Shandong, ChinaPartially supervised learning (PSL) is urgently necessary to explore to construct an efficient universal lesion detection (ULD) segmentation model. An annotated dataset is crucial but hard to acquire because of too many Computed tomography (CT) images and the lack of professionals in computer-aided detection/diagnosis (CADe/CADx). To address this problem, we propose a novel loss function to reduce the proportion of negative anchors which is extremely likely to classify the lesion area (positive samples) as a negative bounding box, further leading to an unexpected performance. Before calculating loss, we generate a mask to intentionally choose fewer negative anchors which will backward wrongful loss to the network. During the process of loss calculation, we set a parameter to reduce the proportion of negative samples, and it significantly reduces the adverse effect of misclassification on the model. Our experiments are implemented in a 3D framework by feeding a partially annotated dataset named DeepLesion, a large-scale public dataset for universal lesion detection from CT. We implement a lot of experiments to choose the most suitable parameter, and the result shows that the proposed method has greatly improved the performance of a ULD detector. Our code can be obtained at https://github.com/PLuld0/PLuldl.https://www.frontiersin.org/articles/10.3389/fphar.2023.1084155/fullPSLULD3D modelmedical image learningneural network learning |
| spellingShingle | Xun Wang Xun Wang Xin Shi Xiangyu Meng Zhiyuan Zhang Chaogang Zhang A universal lesion detection method based on partially supervised learning Frontiers in Pharmacology PSL ULD 3D model medical image learning neural network learning |
| title | A universal lesion detection method based on partially supervised learning |
| title_full | A universal lesion detection method based on partially supervised learning |
| title_fullStr | A universal lesion detection method based on partially supervised learning |
| title_full_unstemmed | A universal lesion detection method based on partially supervised learning |
| title_short | A universal lesion detection method based on partially supervised learning |
| title_sort | universal lesion detection method based on partially supervised learning |
| topic | PSL ULD 3D model medical image learning neural network learning |
| url | https://www.frontiersin.org/articles/10.3389/fphar.2023.1084155/full |
| work_keys_str_mv | AT xunwang auniversallesiondetectionmethodbasedonpartiallysupervisedlearning AT xunwang auniversallesiondetectionmethodbasedonpartiallysupervisedlearning AT xinshi auniversallesiondetectionmethodbasedonpartiallysupervisedlearning AT xiangyumeng auniversallesiondetectionmethodbasedonpartiallysupervisedlearning AT zhiyuanzhang auniversallesiondetectionmethodbasedonpartiallysupervisedlearning AT chaogangzhang auniversallesiondetectionmethodbasedonpartiallysupervisedlearning AT xunwang universallesiondetectionmethodbasedonpartiallysupervisedlearning AT xunwang universallesiondetectionmethodbasedonpartiallysupervisedlearning AT xinshi universallesiondetectionmethodbasedonpartiallysupervisedlearning AT xiangyumeng universallesiondetectionmethodbasedonpartiallysupervisedlearning AT zhiyuanzhang universallesiondetectionmethodbasedonpartiallysupervisedlearning AT chaogangzhang universallesiondetectionmethodbasedonpartiallysupervisedlearning |