Intelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep Learning
Abstract Introduction Compared with traditional fundus examination techniques, ultra-widefield fundus (UWF) images provide 200° panoramic images of the retina, which allows better detection of peripheral retinal lesions. The advent of UWF provides effective solutions only for detection but still lac...
| Main Authors: | , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Adis, Springer Healthcare
2023-01-01
|
| Series: | Ophthalmology and Therapy |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s40123-023-00651-x |
| _version_ | 1827985003981570048 |
|---|---|
| author | Tong Wang Guoliang Liao Lin Chen Yan Zhuang Sibo Zhou Qiongzhen Yuan Lin Han Shanshan Wu Ke Chen Binjian Wang Junyu Mi Yunxia Gao Jiangli Lin Ming Zhang |
| author_facet | Tong Wang Guoliang Liao Lin Chen Yan Zhuang Sibo Zhou Qiongzhen Yuan Lin Han Shanshan Wu Ke Chen Binjian Wang Junyu Mi Yunxia Gao Jiangli Lin Ming Zhang |
| author_sort | Tong Wang |
| collection | DOAJ |
| description | Abstract Introduction Compared with traditional fundus examination techniques, ultra-widefield fundus (UWF) images provide 200° panoramic images of the retina, which allows better detection of peripheral retinal lesions. The advent of UWF provides effective solutions only for detection but still lacks efficient diagnostic capabilities. This study proposed a retinal lesion detection model to automatically locate and identify six relatively typical and high-incidence peripheral retinal lesions from UWF images which will enable early screening and rapid diagnosis. Methods A total of 24,602 augmented ultra-widefield fundus images with labels corresponding to 6 peripheral retinal lesions and normal manifestation labelled by 5 ophthalmologists were included in this study. An object detection model named You Only Look Once X (YOLOX) was modified and trained to locate and classify the six peripheral retinal lesions including rhegmatogenous retinal detachment (RRD), retinal breaks (RB), white without pressure (WWOP), cystic retinal tuft (CRT), lattice degeneration (LD), and paving-stone degeneration (PSD). We applied coordinate attention block and generalized intersection over union (GIOU) loss to YOLOX and evaluated it for accuracy, sensitivity, specificity, precision, F1 score, and average precision (AP). This model was able to show the exact location and saliency map of the retinal lesions detected by the model thus contributing to efficient screening and diagnosis. Results The model reached an average accuracy of 96.64%, sensitivity of 87.97%, specificity of 98.04%, precision of 87.01%, F1 score of 87.39%, and mAP of 86.03% on test dataset 1 including 248 UWF images and reached an average accuracy of 95.04%, sensitivity of 83.90%, specificity of 96.70%, precision of 78.73%, F1 score of 81.96%, and mAP of 80.59% on external test dataset 2 including 586 UWF images, showing this system performs well in distinguishing the six peripheral retinal lesions. Conclusion Focusing on peripheral retinal lesions, this work proposed a deep learning model, which automatically recognized multiple peripheral retinal lesions from UWF images and localized exact positions of lesions. Therefore, it has certain potential for early screening and intelligent diagnosis of peripheral retinal lesions. |
| first_indexed | 2024-04-09T23:07:55Z |
| format | Article |
| id | doaj.art-dbc7f194df874518a8306472d18905eb |
| institution | Directory Open Access Journal |
| issn | 2193-8245 2193-6528 |
| language | English |
| last_indexed | 2024-04-09T23:07:55Z |
| publishDate | 2023-01-01 |
| publisher | Adis, Springer Healthcare |
| record_format | Article |
| series | Ophthalmology and Therapy |
| spelling | doaj.art-dbc7f194df874518a8306472d18905eb2023-03-22T10:36:51ZengAdis, Springer HealthcareOphthalmology and Therapy2193-82452193-65282023-01-011221081109510.1007/s40123-023-00651-xIntelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep LearningTong Wang0Guoliang Liao1Lin Chen2Yan Zhuang3Sibo Zhou4Qiongzhen Yuan5Lin Han6Shanshan Wu7Ke Chen8Binjian Wang9Junyu Mi10Yunxia Gao11Jiangli Lin12Ming Zhang13Department of Ophthalmology, West China HospitalCollege of Biomedical Engineering, Sichuan UniversityDepartment of Ophthalmology, West China HospitalCollege of Biomedical Engineering, Sichuan UniversityCollege of Biomedical Engineering, Sichuan UniversityDepartment of Ophthalmology, West China HospitalCollege of Biomedical Engineering, Sichuan UniversityDepartment of Ophthalmology, West China HospitalCollege of Biomedical Engineering, Sichuan UniversityDepartment of Ophthalmology, West China HospitalCollege of Biomedical Engineering, Sichuan UniversityDepartment of Ophthalmology, West China HospitalCollege of Biomedical Engineering, Sichuan UniversityDepartment of Ophthalmology, West China HospitalAbstract Introduction Compared with traditional fundus examination techniques, ultra-widefield fundus (UWF) images provide 200° panoramic images of the retina, which allows better detection of peripheral retinal lesions. The advent of UWF provides effective solutions only for detection but still lacks efficient diagnostic capabilities. This study proposed a retinal lesion detection model to automatically locate and identify six relatively typical and high-incidence peripheral retinal lesions from UWF images which will enable early screening and rapid diagnosis. Methods A total of 24,602 augmented ultra-widefield fundus images with labels corresponding to 6 peripheral retinal lesions and normal manifestation labelled by 5 ophthalmologists were included in this study. An object detection model named You Only Look Once X (YOLOX) was modified and trained to locate and classify the six peripheral retinal lesions including rhegmatogenous retinal detachment (RRD), retinal breaks (RB), white without pressure (WWOP), cystic retinal tuft (CRT), lattice degeneration (LD), and paving-stone degeneration (PSD). We applied coordinate attention block and generalized intersection over union (GIOU) loss to YOLOX and evaluated it for accuracy, sensitivity, specificity, precision, F1 score, and average precision (AP). This model was able to show the exact location and saliency map of the retinal lesions detected by the model thus contributing to efficient screening and diagnosis. Results The model reached an average accuracy of 96.64%, sensitivity of 87.97%, specificity of 98.04%, precision of 87.01%, F1 score of 87.39%, and mAP of 86.03% on test dataset 1 including 248 UWF images and reached an average accuracy of 95.04%, sensitivity of 83.90%, specificity of 96.70%, precision of 78.73%, F1 score of 81.96%, and mAP of 80.59% on external test dataset 2 including 586 UWF images, showing this system performs well in distinguishing the six peripheral retinal lesions. Conclusion Focusing on peripheral retinal lesions, this work proposed a deep learning model, which automatically recognized multiple peripheral retinal lesions from UWF images and localized exact positions of lesions. Therefore, it has certain potential for early screening and intelligent diagnosis of peripheral retinal lesions.https://doi.org/10.1007/s40123-023-00651-xPeripheral retinal lesionUltra-widefield fundusDeep learningObject detectionYou Only Look Once X |
| spellingShingle | Tong Wang Guoliang Liao Lin Chen Yan Zhuang Sibo Zhou Qiongzhen Yuan Lin Han Shanshan Wu Ke Chen Binjian Wang Junyu Mi Yunxia Gao Jiangli Lin Ming Zhang Intelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep Learning Ophthalmology and Therapy Peripheral retinal lesion Ultra-widefield fundus Deep learning Object detection You Only Look Once X |
| title | Intelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep Learning |
| title_full | Intelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep Learning |
| title_fullStr | Intelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep Learning |
| title_full_unstemmed | Intelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep Learning |
| title_short | Intelligent Diagnosis of Multiple Peripheral Retinal Lesions in Ultra-widefield Fundus Images Based on Deep Learning |
| title_sort | intelligent diagnosis of multiple peripheral retinal lesions in ultra widefield fundus images based on deep learning |
| topic | Peripheral retinal lesion Ultra-widefield fundus Deep learning Object detection You Only Look Once X |
| url | https://doi.org/10.1007/s40123-023-00651-x |
| work_keys_str_mv | AT tongwang intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT guoliangliao intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT linchen intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT yanzhuang intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT sibozhou intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT qiongzhenyuan intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT linhan intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT shanshanwu intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT kechen intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT binjianwang intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT junyumi intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT yunxiagao intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT jianglilin intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning AT mingzhang intelligentdiagnosisofmultipleperipheralretinallesionsinultrawidefieldfundusimagesbasedondeeplearning |