Xiaoling Xie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xiaoling Xie has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
- Ce Zheng
Department of Ophthalmology, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; Joint Shantou International Eye Center, Shantou University, Chinese University of Hong Kong, Shantou University Medical College, Shantou, China
3 shared publications - Hongfei Ye
Ophthalmology, Xinhua Hospital affiliated to Shanghai Jiaotong University School of Medicine, Shanghai, China
3 shared publications - Jianlong Yang
School of Biomedical Engineering Shanghai Jiao Tong University Shanghai China
3 shared publications - Peiquan Zhao
Department of Ophthalmology, Xinhua Hospital affiliated to Shanghai Jiao Tong University School of Medicine
3 shared publications - Jili Chen
Department of Ophthalmology, Shibei Hospital, Shanghai, China
2 shared publications - Mingzhi Zhang
Joint Shantou International Eye Center of Shantou University and the Chinese University of Hong Kong, Shantou, Guangdong, China
2 shared publications
Latest Publications
- 3 March 2026
- Published by Springer Nature in Nature Biomedical Engineering
An ophthalmic video foundation model for surgical recognition and navigation with wet-lab porcine eye validation
Foundation models in artificial intelligence are revolutionizing healthcare by utilizing large-scale unlabelled data for pretraining. However, their intraoperative applications remain underexplored owing to limited surgical data and the challenges of real-time deployment. Here we show the development of the ophthalmic video foundation model (OVFM), designed for microscopic ophthalmic surgical recognition and navigation. Leveraging a self-supervised video transformer structure and trained on an ophthalmic video dataset comprising 1.1 million clips across 144 surgical types, OVFM learns the spatiotemporal motion features of ophthalmic procedures. We demonstrate OVFM's superior performance across seven downstream tasks. To enable real-time use, we applied knowledge distillation, reducing the model's size while retaining its accuracy, which allows for deployment on surgical microscope units. In cataract surgeries performed by ten surgeons on wet-lab porcine eyes, the OVFM-powered system enhanced surgical performance and reduced skill gaps, demonstrating notable potential for real-time, intraoperative applications across various surgical fields.
- 3 March 2026
- Published by Springer Nature in Nature Biomedical Engineering
An ophthalmic video foundation model for surgical recognition and navigation with wet-lab porcine eye validation
Foundation models in artificial intelligence are revolutionizing healthcare by utilizing large-scale unlabelled data for pretraining. However, their intraoperative applications remain underexplored owing to limited surgical data and the challenges of real-time deployment. Here we show the development of the ophthalmic video foundation model (OVFM), designed for microscopic ophthalmic surgical recognition and navigation. Leveraging a self-supervised video transformer structure and trained on an ophthalmic video dataset comprising 1.1 million clips across 144 surgical types, OVFM learns the spatiotemporal motion features of ophthalmic procedures. We demonstrate OVFM's superior performance across seven downstream tasks. To enable real-time use, we applied knowledge distillation, reducing the model's size while retaining its accuracy, which allows for deployment on surgical microscope units. In cataract surgeries performed by ten surgeons on wet-lab porcine eyes, the OVFM-powered system enhanced surgical performance and reduced skill gaps, demonstrating notable potential for real-time, intraoperative applications across various surgical fields.
- Open Access
- 17 July 2024
- Published by BMJ in British Journal of Ophthalmology
- Vol. 108 (10) , 1390-1397
Development and evaluation of a large language model of ophthalmology in Chinese
Background Large language models (LLMs), such as ChatGPT, have considerable implications for various medical applications. However, ChatGPT’s training primarily draws from English-centric internet data and is not tailored explicitly to the medical domain. Thus, an ophthalmic LLM in Chinese is clinically essential for both healthcare providers and patients in mainland China. Methods We developed an LLM of ophthalmology (MOPH) using Chinese corpora and evaluated its performance in three clinical scenarios: ophthalmic board exams in Chinese, answering evidence-based medicine-oriented ophthalmic questions and diagnostic accuracy for clinical vignettes. Additionally, we compared MOPH’s performance to that of human doctors. Results In the ophthalmic exam, MOPH’s average score closely aligned with the mean score of trainees (64.7 (range 62–68) vs 66.2 (range 50–92), p=0.817), but achieving a score above 60 in all seven mock exams. In answering ophthalmic questions, MOPH demonstrated an adherence of 83.3% (25/30) of responses following Chinese guidelines (Likert scale 4–5). Only 6.7% (2/30, Likert scale 1–2) and 10% (3/30, Likert scale 3) of responses were rated as ‘poor or very poor’ or ‘potentially misinterpretable inaccuracies’ by reviewers. In diagnostic accuracy, although the rate of correct diagnosis by ophthalmologists was superior to that by MOPH (96.1% vs 81.1%, p>0.05), the difference was not statistically significant. Conclusion This study demonstrated the promising performance of MOPH, a Chinese-specific ophthalmic LLM, in diverse clinical scenarios. MOPH has potential real-world applications in Chinese-language ophthalmology settings.
- 17 July 2024
- Published by BMJ in British Journal of Ophthalmology
- Vol. 108 (10) , 1390-1397
Open AccessDevelopment and evaluation of a large language model of ophthalmology in Chinese
Background Large language models (LLMs), such as ChatGPT, have considerable implications for various medical applications. However, ChatGPT’s training primarily draws from English-centric internet data and is not tailored explicitly to the medical domain. Thus, an ophthalmic LLM in Chinese is clinically essential for both healthcare providers and patients in mainland China. Methods We developed an LLM of ophthalmology (MOPH) using Chinese corpora and evaluated its performance in three clinical scenarios: ophthalmic board exams in Chinese, answering evidence-based medicine-oriented ophthalmic questions and diagnostic accuracy for clinical vignettes. Additionally, we compared MOPH’s performance to that of human doctors. Results In the ophthalmic exam, MOPH’s average score closely aligned with the mean score of trainees (64.7 (range 62–68) vs 66.2 (range 50–92), p=0.817), but achieving a score above 60 in all seven mock exams. In answering ophthalmic questions, MOPH demonstrated an adherence of 83.3% (25/30) of responses following Chinese guidelines (Likert scale 4–5). Only 6.7% (2/30, Likert scale 1–2) and 10% (3/30, Likert scale 3) of responses were rated as ‘poor or very poor’ or ‘potentially misinterpretable inaccuracies’ by reviewers. In diagnostic accuracy, although the rate of correct diagnosis by ophthalmologists was superior to that by MOPH (96.1% vs 81.1%, p>0.05), the difference was not statistically significant. Conclusion This study demonstrated the promising performance of MOPH, a Chinese-specific ophthalmic LLM, in diverse clinical scenarios. MOPH has potential real-world applications in Chinese-language ophthalmology settings.
- Open Access
- 7 January 2024
- Published by JMIR Publications Inc. in JMIR Preprints
Generative Adversarial Networks Synthetic Optical Coherence Tomography Images as an Education Tool for Image Diagnosis of Macular Diseases: A Randomized Trial (Preprint)
BACKGROUND Macular diseases are one of the leading causes of blindness. Diagnosis of macular diseases is an important part of ophthalmic education. Optical coherence tomography (OCT) is now a standard imaging modality for diagnosing and treating macular disease. Accurate interpretations of OCT with macular diseases are vital for ophthalmic resident training worldwide. Previously, we proposed generative adversarial networks (GANs) model for synthetic OCT images of macular disorders. Our result also showed its potential value for educational purposes. This study aimed to establish an educational tool based on GANs synthetic images for ophthalmic residents’ and medical students’ education and evaluate its validity. OBJECTIVE To evaluate the assessment of GANs synthetic OCT images as an education tool for image diagnosis of macular diseases in medical students and ophthalmic residents trainingIn Group GANs OCT, scores improved from 80.0(46.0, 85.5) to 92.0(81.0,95.5) five minutes after training (p<0.001) and 92.30± 5.36 one-week after training (p<0.001). Similarly, in Group real OCT, scores improved from 66.00±19.52 to 92.90± 5.71(p<0.001) and 93.20± 6.10 (p<0.001), respectively. No significantly difference were found for scores of tests, score improvements and time consumptions between Group GANs OCT and Group real OCT. Medical students had significantly higher score improvement after training than residents(p<0.001) did. METHODS In this randomized trial, twenty fifth-year medical students and 20 ophthalmic residents were enrolled and randomly assigned (1:1 allocation) into Group real OCT and Group GANs OCT. All participants had a pre-test to show their educational background, followed by a 30-minute-smartphone-based GANs/real OCT images-based education program for macular diseases recognition training. Another two tests were scheduled five minutes and one week after the training to test their short-term and long-term performances. Scores and time consumptions were recorded and compared. After all tests, an anonymous subjective questionnaire was acquired. RESULTS In Group GANs OCT, scores improved from 80.0(46.0, 85.5) to 92.0(81.0,95.5) five minutes after training (p<0.001) and 92.30± 5.36 one-week after training (p<0.001). Similarly, in Group real OCT, scores improved from 66.00±19.52 to 92.90± 5.71(p<0.001) and 93.20± 6.10 (p<0.001), respectively. No significantly difference were found for scores of tests, score improvements and time consumptions between Group GANs OCT and Group real OCT. Medical students had significantly higher score improvement after training than residents(p<0.001) did. CONCLUSIONS The education tool using synthetic OCT images had a similar educational ability compared to that using real OCT images, which improved the interpretation ability of ophthalmic residents and medical students in both short-term and long-term performances. The smartphone-based educational tool could be promoted more widely for educational applications. CLINICALTRIAL Chinses Clinical Trial Registry, No. ChiCTR 2100053195.
- 7 January 2024
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessGenerative Adversarial Networks Synthetic Optical Coherence Tomography Images as an Education Tool for Image Diagnosis of Macular Diseases: A Randomized Trial (Preprint)
BACKGROUND Macular diseases are one of the leading causes of blindness. Diagnosis of macular diseases is an important part of ophthalmic education. Optical coherence tomography (OCT) is now a standard imaging modality for diagnosing and treating macular disease. Accurate interpretations of OCT with macular diseases are vital for ophthalmic resident training worldwide. Previously, we proposed generative adversarial networks (GANs) model for synthetic OCT images of macular disorders. Our result also showed its potential value for educational purposes. This study aimed to establish an educational tool based on GANs synthetic images for ophthalmic residents’ and medical students’ education and evaluate its validity. OBJECTIVE To evaluate the assessment of GANs synthetic OCT images as an education tool for image diagnosis of macular diseases in medical students and ophthalmic residents trainingIn Group GANs OCT, scores improved from 80.0(46.0, 85.5) to 92.0(81.0,95.5) five minutes after training (p<0.001) and 92.30± 5.36 one-week after training (p<0.001). Similarly, in Group real OCT, scores improved from 66.00±19.52 to 92.90± 5.71(p<0.001) and 93.20± 6.10 (p<0.001), respectively. No significantly difference were found for scores of tests, score improvements and time consumptions between Group GANs OCT and Group real OCT. Medical students had significantly higher score improvement after training than residents(p<0.001) did. METHODS In this randomized trial, twenty fifth-year medical students and 20 ophthalmic residents were enrolled and randomly assigned (1:1 allocation) into Group real OCT and Group GANs OCT. All participants had a pre-test to show their educational background, followed by a 30-minute-smartphone-based GANs/real OCT images-based education program for macular diseases recognition training. Another two tests were scheduled five minutes and one week after the training to test their short-term and long-term performances. Scores and time consumptions were recorded and compared. After all tests, an anonymous subjective questionnaire was acquired. RESULTS In Group GANs OCT, scores improved from 80.0(46.0, 85.5) to 92.0(81.0,95.5) five minutes after training (p<0.001) and 92.30± 5.36 one-week after training (p<0.001). Similarly, in Group real OCT, scores improved from 66.00±19.52 to 92.90± 5.71(p<0.001) and 93.20± 6.10 (p<0.001), respectively. No significantly difference were found for scores of tests, score improvements and time consumptions between Group GANs OCT and Group real OCT. Medical students had significantly higher score improvement after training than residents(p<0.001) did. CONCLUSIONS The education tool using synthetic OCT images had a similar educational ability compared to that using real OCT images, which improved the interpretation ability of ophthalmic residents and medical students in both short-term and long-term performances. The smartphone-based educational tool could be promoted more widely for educational applications. CLINICALTRIAL Chinses Clinical Trial Registry, No. ChiCTR 2100053195.
- Open Access
- 22 September 2021
- Published by Springer Nature in BMC Ophthalmology
- Vol. 21 (1) , 1-9
Detection of shallow anterior chamber depth from two-dimensional anterior segment photographs using deep learning
Background: The purpose of this study was to implement and evaluate a deep learning (DL) approach for automatically detecting shallow anterior chamber depth (ACD) from two-dimensional (2D) overview anterior segment photographs. Methods: We trained a DL model using a dataset of anterior segment photographs collected from Shanghai Aier Eye Hospital from June 2018 to December 2019. A Pentacam HR system was used to capture a 2D overview eye image and measure the ACD. Shallow ACD was defined as ACD less than 2.4 mm. The DL model was evaluated by a five-fold cross-validation test in a hold-out testing dataset. We also evaluated the DL model by testing it against two glaucoma specialists. The performance of the DL model was calculated by metrics, including accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results: A total of 3753 photographs (1720 shallow AC and 2033 deep AC images) were assigned to the training dataset, and 1302 photographs (509 shallow AC and 793 deep AC images) were held out for two internal testing datasets. In detecting shallow ACD in the internal hold-out testing dataset, the DL model achieved an AUC of 0.86 (95% CI, 0.83–0.90) with 80% sensitivity and 79% specificity. In the same testing dataset, the DL model also achieved better performance than the two glaucoma specialists (accuracy of 80% vs. accuracy of 74 and 69%). Conclusions: We proposed a high-performing DL model to automatically detect shallow ACD from overview anterior segment photographs. Our DL model has potential applications in detecting and monitoring shallow ACD in the real world. Trial registration: http://clinicaltrials.gov, NCT04340635, retrospectively registered on 29 March 2020.
- 22 September 2021
- Published by Springer Nature in BMC Ophthalmology
- Vol. 21 (1) , 1-9
Open AccessDetection of shallow anterior chamber depth from two-dimensional anterior segment photographs using deep learning
Background: The purpose of this study was to implement and evaluate a deep learning (DL) approach for automatically detecting shallow anterior chamber depth (ACD) from two-dimensional (2D) overview anterior segment photographs. Methods: We trained a DL model using a dataset of anterior segment photographs collected from Shanghai Aier Eye Hospital from June 2018 to December 2019. A Pentacam HR system was used to capture a 2D overview eye image and measure the ACD. Shallow ACD was defined as ACD less than 2.4 mm. The DL model was evaluated by a five-fold cross-validation test in a hold-out testing dataset. We also evaluated the DL model by testing it against two glaucoma specialists. The performance of the DL model was calculated by metrics, including accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results: A total of 3753 photographs (1720 shallow AC and 2033 deep AC images) were assigned to the training dataset, and 1302 photographs (509 shallow AC and 793 deep AC images) were held out for two internal testing datasets. In detecting shallow ACD in the internal hold-out testing dataset, the DL model achieved an AUC of 0.86 (95% CI, 0.83–0.90) with 80% sensitivity and 79% specificity. In the same testing dataset, the DL model also achieved better performance than the two glaucoma specialists (accuracy of 80% vs. accuracy of 74 and 69%). Conclusions: We proposed a high-performing DL model to automatically detect shallow ACD from overview anterior segment photographs. Our DL model has potential applications in detecting and monitoring shallow ACD in the real world. Trial registration: http://clinicaltrials.gov, NCT04340635, retrospectively registered on 29 March 2020.
- Open Access
Assessment of Generative Adversarial Networks Model for Synthetic Optical Coherence Tomography Images of Retinal Disorders
Purpose: To assess whether a generative adversarial network (GAN) could synthesize realistic optical coherence tomography (OCT) images that satisfactorily serve as the educational images for retinal specialists and the training datasets for the classification of various retinal disorders using deep learning (DL). Methods: The GANs architecture was adopted to synthesis high-resolution OCT images training on a publicly available OCT dataset including urgent referrals (choroidal neovascularization and diabetic macular edema) and non-urgent referrals (normal and drusen). 400 real and synthetic OCT images were evaluated by 2 retinal specialists to assess image quality. We further trained 2 DL models on either real or synthetic datasets and compared the performance of urgent vs nonurgent referrals diagnosis tested on a local (1000 images from the public dataset) and clinical validation dataset (278 images from Shanghai Shibei Hospital). Results: The image quality of real vs synthetic OCT images was similar as assessed by 2 retinal specialists. The accuracy of discrimination as real vs synthetic OCT images was 59.50% for retinal specialist 1 and 53.67% for retinal specialist 2. For the local dataset, the DL model trained on real (DL_Model_R) and synthetic OCT images (DL_Model_S) had an area under the curve (AUC) of 0.99, and 0.98 respectively. For the clinical dataset, the AUC was 0.94 for DL_Model_R, 0.90 for DL_Model_S. Conclusions: The GAN-synthetic OCT images can be used by clinicians for educational purposes and developing DL algorithms. Translational Relevance: The medical image synthesis based on GANs is promising in human and machine to fulfill clinical tasks.
Open AccessAssessment of Generative Adversarial Networks Model for Synthetic Optical Coherence Tomography Images of Retinal Disorders
Purpose: To assess whether a generative adversarial network (GAN) could synthesize realistic optical coherence tomography (OCT) images that satisfactorily serve as the educational images for retinal specialists and the training datasets for the classification of various retinal disorders using deep learning (DL). Methods: The GANs architecture was adopted to synthesis high-resolution OCT images training on a publicly available OCT dataset including urgent referrals (choroidal neovascularization and diabetic macular edema) and non-urgent referrals (normal and drusen). 400 real and synthetic OCT images were evaluated by 2 retinal specialists to assess image quality. We further trained 2 DL models on either real or synthetic datasets and compared the performance of urgent vs nonurgent referrals diagnosis tested on a local (1000 images from the public dataset) and clinical validation dataset (278 images from Shanghai Shibei Hospital). Results: The image quality of real vs synthetic OCT images was similar as assessed by 2 retinal specialists. The accuracy of discrimination as real vs synthetic OCT images was 59.50% for retinal specialist 1 and 53.67% for retinal specialist 2. For the local dataset, the DL model trained on real (DL_Model_R) and synthetic OCT images (DL_Model_S) had an area under the curve (AUC) of 0.99, and 0.98 respectively. For the clinical dataset, the AUC was 0.94 for DL_Model_R, 0.90 for DL_Model_S. Conclusions: The GAN-synthetic OCT images can be used by clinicians for educational purposes and developing DL algorithms. Translational Relevance: The medical image synthesis based on GANs is promising in human and machine to fulfill clinical tasks.