loading page

Differentiation of eosinophilic and non-eosinophilic chronic rhinosinusitis on preoperative computed tomography using deep learning
  • +5
  • Hongli Hua,
  • Song Li,
  • Yu Xu,
  • Shiming Chen,
  • Yonggang Kong,
  • Rui Yang,
  • Yu-Qin Deng,
  • Zezhang Tao
Hongli Hua
Renmin Hospital of Wuhan University

Corresponding Author:[email protected]

Author Profile
Song Li
Renmin Hospital of Wuhan University
Author Profile
Yu Xu
Renmin Hospital of Wuhan University
Author Profile
Shiming Chen
Renmin Hospital of Wuhan University
Author Profile
Yonggang Kong
Renmin Hospital of Wuhan University
Author Profile
Rui Yang
Renmin Hospital of Wuhan University
Author Profile
Yu-Qin Deng
Renmin Hospital of Wuhan University
Author Profile
Zezhang Tao
Renmin Hospital of Wuhan University
Author Profile

Abstract

Objective: This study aimed to develop deep learning (DL) models for differentiating between eosinophilic chronic rhinosinusitis (ECRS) and non-eosinophilic chronic rhinosinusitis (NECRS) on preoperative computed tomography (CT). Methods: A total of 878 chronic rhinosinusitis (CRS) patients undergoing nasal endoscopic surgery were included. Axial spiral CT images were pre-processed and used to build the dataset. Two semantic segmentation models based on U-net and Deeplabv3 were trained to segment sinus area in CT images. All patient images were segmented using the better-performing segmentation model and used for training and validation of the transferred efficientnet_b0, resnet50, inception_resnet_v2, and Xception neural networks. Additionally, we evaluated the performances of the models trained using each image and each patient as a unit. The precision of each model was assessed based on the receiver operating characteristic curve. Further, we analyzed the confusion matrix, accuracy, and interpretability of each model. Results: The Dice coefficients of U-net and Deeplabv3 were 0.953 and 0.961, respectively. The average area under the curve and mean accuracy values of the four networks were 0.848 and 0.762 for models trained using a single image as a unit, while the corresponding values for models trained using each patient as a unit were 0.853 and 0.893, respectively. The generated Grad-Cams showed good interpretability. Conclusion: Combining semantic segmentation with classification networks could effectively distinguish between patients with ECRS and NECRS based on preoperative sinus CT images. Furthermore, labeling each patient to build a dataset for classification may be more reliable than labeling each medical image.
16 Mar 2022Submitted to Clinical Otolaryngology
12 Apr 2022Submission Checks Completed
12 Apr 2022Assigned to Editor
19 Apr 2022Reviewer(s) Assigned
02 May 2022Review(s) Completed, Editorial Evaluation Pending
14 May 2022Editorial Decision: Revise Major
14 Jun 20221st Revision Received
28 Jun 2022Submission Checks Completed
28 Jun 2022Assigned to Editor
29 Jun 2022Reviewer(s) Assigned
12 Jul 2022Review(s) Completed, Editorial Evaluation Pending
21 Jul 2022Editorial Decision: Revise Minor
10 Aug 20222nd Revision Received
17 Aug 2022Submission Checks Completed
17 Aug 2022Assigned to Editor
20 Aug 2022Reviewer(s) Assigned
11 Sep 2022Review(s) Completed, Editorial Evaluation Pending
11 Sep 2022Editorial Decision: Accept
14 Oct 2022Published in Clinical Otolaryngology. 10.1111/coa.13988