Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic Images
Chronic rhinosinusitis with nasal polyps (CRSwNP), as one of the most common chronic nasal inflammations, has been a major research focus in the field of rhinology due to its complex and varied pathophysiological features and suboptimal clinical treatment outcomes. However, existing diagnostic metho...
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10845750/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832583207603666944 |
|---|---|
| author | Yuhang Wang Xiaolei Zhang Weidong Du Na Dai Yi Lyv Keying Wu Yiyang Tian Yuxin Jie Yu Lin Weipiao Kang |
| author_facet | Yuhang Wang Xiaolei Zhang Weidong Du Na Dai Yi Lyv Keying Wu Yiyang Tian Yuxin Jie Yu Lin Weipiao Kang |
| author_sort | Yuhang Wang |
| collection | DOAJ |
| description | Chronic rhinosinusitis with nasal polyps (CRSwNP), as one of the most common chronic nasal inflammations, has been a major research focus in the field of rhinology due to its complex and varied pathophysiological features and suboptimal clinical treatment outcomes. However, existing diagnostic methods still face many challenges, especially shortcomings in accurate typing and the development of individualized treatment plans. Currently, sinus CT is an essential non-invasive tool for preoperative assessment of CRSwNP endotypes, and formulation of personalized and precise treatment strategies. It provides otolaryngologists with a valuable means to successfully diagnose and treat CRSwNP patients. This paper introduces a 3D semantic segmentation technology to achieve fully automatic 3D segmentation of sinus lesion regions, allowing physicians to observe the anatomical structures and lesions in the sinuses more clearly, thereby improving the diagnostic accuracy for CRSwNP. The study involved 242 Computed Tomography (CT) images of patients with CRSwNP, constructing a high-quality professional CRSwNP database for the training, validation, and testing of neural networks. We chose a custom 3D nnU-Net v2 network model because of its excellent performance in the field of 3D medical image segmentation, especially in automated training and accurate segmentation of complex structures. The model can accurately segment the sinus cavity by deeply learning the microstructure and deep features of 3D sinus CT images. Testing results demonstrated that the model accurately identified the segmentation areas, achieving a Dice Similarity Coefficient of 92.8%, Intersection over Union of 86.64%, accuracy of 99.69%, precision of 92.63%, and recall of 93.22%. This deep learning-based fully automatic CRSwNP sinus segmentation model exhibits excellent segmentation performance, aiding clinicians in further diagnosing CRSwNP endotypes and contributing to the advancement of clinical application deployment. |
| format | Article |
| id | doaj-art-11f3aad101ff45b0befa99d9fd8e18e2 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-11f3aad101ff45b0befa99d9fd8e18e22025-01-29T00:01:01ZengIEEEIEEE Access2169-35362025-01-0113164441645410.1109/ACCESS.2025.353139610845750Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic ImagesYuhang Wang0Xiaolei Zhang1https://orcid.org/0000-0002-5909-5820Weidong Du2Na Dai3Yi Lyv4Keying Wu5Yiyang Tian6Yuxin Jie7Yu Lin8Weipiao Kang9Department of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Radiology, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaDepartment of Otolaryngology-Head and Neck Surgery, Second Affiliated Hospital of Shantou University Medical College, Shantou, ChinaChronic rhinosinusitis with nasal polyps (CRSwNP), as one of the most common chronic nasal inflammations, has been a major research focus in the field of rhinology due to its complex and varied pathophysiological features and suboptimal clinical treatment outcomes. However, existing diagnostic methods still face many challenges, especially shortcomings in accurate typing and the development of individualized treatment plans. Currently, sinus CT is an essential non-invasive tool for preoperative assessment of CRSwNP endotypes, and formulation of personalized and precise treatment strategies. It provides otolaryngologists with a valuable means to successfully diagnose and treat CRSwNP patients. This paper introduces a 3D semantic segmentation technology to achieve fully automatic 3D segmentation of sinus lesion regions, allowing physicians to observe the anatomical structures and lesions in the sinuses more clearly, thereby improving the diagnostic accuracy for CRSwNP. The study involved 242 Computed Tomography (CT) images of patients with CRSwNP, constructing a high-quality professional CRSwNP database for the training, validation, and testing of neural networks. We chose a custom 3D nnU-Net v2 network model because of its excellent performance in the field of 3D medical image segmentation, especially in automated training and accurate segmentation of complex structures. The model can accurately segment the sinus cavity by deeply learning the microstructure and deep features of 3D sinus CT images. Testing results demonstrated that the model accurately identified the segmentation areas, achieving a Dice Similarity Coefficient of 92.8%, Intersection over Union of 86.64%, accuracy of 99.69%, precision of 92.63%, and recall of 93.22%. This deep learning-based fully automatic CRSwNP sinus segmentation model exhibits excellent segmentation performance, aiding clinicians in further diagnosing CRSwNP endotypes and contributing to the advancement of clinical application deployment.https://ieeexplore.ieee.org/document/10845750/Chronic sinusitisconvolutional neural networkdeep learningparanasal sinusessegmentation |
| spellingShingle | Yuhang Wang Xiaolei Zhang Weidong Du Na Dai Yi Lyv Keying Wu Yiyang Tian Yuxin Jie Yu Lin Weipiao Kang Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic Images IEEE Access Chronic sinusitis convolutional neural network deep learning paranasal sinuses segmentation |
| title | Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic Images |
| title_full | Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic Images |
| title_fullStr | Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic Images |
| title_full_unstemmed | Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic Images |
| title_short | Deep Learning-Based Fully Automatic Segmentation of the Paranasal Sinuses in Chronic Rhinosinusitis Patients Using Computed Tomographic Images |
| title_sort | deep learning based fully automatic segmentation of the paranasal sinuses in chronic rhinosinusitis patients using computed tomographic images |
| topic | Chronic sinusitis convolutional neural network deep learning paranasal sinuses segmentation |
| url | https://ieeexplore.ieee.org/document/10845750/ |
| work_keys_str_mv | AT yuhangwang deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT xiaoleizhang deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT weidongdu deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT nadai deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT yilyv deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT keyingwu deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT yiyangtian deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT yuxinjie deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT yulin deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages AT weipiaokang deeplearningbasedfullyautomaticsegmentationoftheparanasalsinusesinchronicrhinosinusitispatientsusingcomputedtomographicimages |