Junbo Zeng1, Wenting Deng1,2, Jingang Yu3, Lichao Xiao3, Suijun Chen1, Xueyuan Zhang1, Linqi Zeng4, Donglang Chen4, Peng Li5, Yubin Chen5, Hongzheng Zhang6, Fan Shu6, Minjian Wu1, Yuejia Su1, Yuanqing Li2, Yuexin Cai7,8, Yiqing Zheng9,10. 1. Department of Otolaryngology, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, 107# Yanjiang West Road, Guangzhou, China. 2. Shenshan Medical Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Shanwei, Guangdong, China. 3. School of Automation Science and Engineering, South China University of Technology, Guangzhou, China. 4. Zhongshan School of Medicine, Sun Yat-Sen University, Guangzhou, China. 5. Department of Otolaryngology Head and Neck Surgery, The Third Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China. 6. Department of Otolaryngology-Head and Neck Surgery, Zhujiang Hospital, South Medical University, Guangzhou, China. 7. Department of Otolaryngology, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, 107# Yanjiang West Road, Guangzhou, China. caiyx25@mail.sysu.edu.cn. 8. Shenshan Medical Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Shanwei, Guangdong, China. caiyx25@mail.sysu.edu.cn. 9. Department of Otolaryngology, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, 107# Yanjiang West Road, Guangzhou, China. zhengyiq@mail.sysu.edu.cn. 10. Shenshan Medical Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, Shanwei, Guangdong, China. zhengyiq@mail.sysu.edu.cn.
Abstract
BACKGROUND: This study aimed to develop and validate a deep learning (DL) model to identify atelectasis and attic retraction pocket in cases of otitis media with effusion (OME) using multi-center otoscopic images. METHOD: A total of 6393 OME otoscopic images from three centers were used to develop and validate a DL model for detecting atelectasis and attic retraction pocket. A threefold random cross-validation procedure was adopted to divide the dataset into training validation sets on a patient level. A team of otologists was assigned to diagnose and characterize atelectasis and attic retraction pocket in otoscopic images. Receiver operating characteristic (ROC) curves, including area under the ROC curve (AUC), accuracy, sensitivity, and specificity were used to assess the performance of the DL model. Class Activation Mapping (CAM) illustrated the discriminative regions in the otoscopic images. RESULTS: Among all OME otoscopic images, 3564 (55.74%) were identified with attic retraction pocket, and 2460 (38.48%) with atelectasis. The diagnostic DL model of attic retraction pocket and atelectasis achieved a threefold cross-validation accuracy of 89% and 79%, AUC of 0.89 and 0.87, a sensitivity of 0.93 and 0.71, and a specificity of 0.62 and 0.84, respectively. Larger and deeper cases of atelectasis and attic retraction pocket showed greater weight, based on the red color depicted in the heat map of CAM. CONCLUSION: The DL algorithm could be employed to identify atelectasis and attic retraction pocket in otoscopic images of OME, and as a tool to assist in the accurate diagnosis of OME.
BACKGROUND: This study aimed to develop and validate a deep learning (DL) model to identify atelectasis and attic retraction pocket in cases of otitis media with effusion (OME) using multi-center otoscopic images. METHOD: A total of 6393 OME otoscopic images from three centers were used to develop and validate a DL model for detecting atelectasis and attic retraction pocket. A threefold random cross-validation procedure was adopted to divide the dataset into training validation sets on a patient level. A team of otologists was assigned to diagnose and characterize atelectasis and attic retraction pocket in otoscopic images. Receiver operating characteristic (ROC) curves, including area under the ROC curve (AUC), accuracy, sensitivity, and specificity were used to assess the performance of the DL model. Class Activation Mapping (CAM) illustrated the discriminative regions in the otoscopic images. RESULTS: Among all OME otoscopic images, 3564 (55.74%) were identified with attic retraction pocket, and 2460 (38.48%) with atelectasis. The diagnostic DL model of attic retraction pocket and atelectasis achieved a threefold cross-validation accuracy of 89% and 79%, AUC of 0.89 and 0.87, a sensitivity of 0.93 and 0.71, and a specificity of 0.62 and 0.84, respectively. Larger and deeper cases of atelectasis and attic retraction pocket showed greater weight, based on the red color depicted in the heat map of CAM. CONCLUSION: The DL algorithm could be employed to identify atelectasis and attic retraction pocket in otoscopic images of OME, and as a tool to assist in the accurate diagnosis of OME.
Authors: Richard M Rosenfeld; Jennifer J Shin; Seth R Schwartz; Robyn Coggins; Lisa Gagnon; Jesse M Hackell; David Hoelting; Lisa L Hunter; Ann W Kummer; Spencer C Payne; Dennis S Poe; Maria Veling; Peter M Vila; Sandra A Walsh; Maureen D Corrigan Journal: Otolaryngol Head Neck Surg Date: 2016-02 Impact factor: 3.497