A hierarchical deep learning approach with transparency and interpretability based on small samples for glaucoma diagnosis

Xu, Yongli; Hu, Man; Liu, Hanruo; Yang, Hao; Wang, Huaizhou; Lu, Shiping; Liang, Tao; Li, Xiaoxing; Xu, Mai; Li, Liu; Li, Huiqi; Ji, Xin; Wang, Zhijun; Li, Li; Weinreb, Robert N.; Wang, Ke

doi:10.1038/s41746-021-00417-4

Cited by 28 publications

(22 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The validation set of this study is a subset randomly selected from the extensive sample dataset of our previous study [ 22 ], including 500 cross-sectional (one image per patient at every time point) fundus images of glaucoma and 500 cross-sectional fundus images of normal eyes from Beijing Tongren Hospital. Eyes were diagnosed with glaucoma if there were both glaucomatous optic neuropathy (GON) and glaucomatous VF defects.…”

Section: Methodsmentioning

confidence: 99%

“…Glaucomatous VF loss was diagnosed if any of the following findings were evident on two consecutive VF tests: a glaucoma hemifield test outside normal limits, pattern standard deviation (PSD) < 5%, or a cluster of three or more nonedged points in typical glaucomatous locations, all depressed on the pattern deviation plot at a level of p < 0.05, with one point in the cluster depressed at a level of p < 0.01 [ 23 ]. According to the diagnosis results of these fundus images by the AI system [ 22 ], the verification set contains 480 true positives, 20 false negatives, 460 true negatives, and 40 false positives. The entire validation set was randomly divided into five groups.…”

Section: Methodsmentioning

confidence: 99%

“…Previously, we established a DL system with a hierarchical structure (HDLS) based on a small number of samples [ 22 ]. This HDLS can comprehensively simulate the diagnostic thinking of human experts.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Practical Application of Artificial Intelligence Technology in Glaucoma Diagnosis

Gong

Yin

et al. 2022

Journal of Ophthalmology

Self Cite

View full text Add to dashboard Cite

Purpose. By comparing the performance of different models between artificial intelligence (AI) and doctors, we aim to evaluate and identify the optimal model for future usage of AI. Methods. A total of 500 fundus images of glaucoma and 500 fundus images of normal eyes were collected and randomly divided into five groups, with each group corresponding to one round. The AI system provided diagnostic suggestions for each image. Four doctors provided diagnoses without the assistance of the AI in the first round and with the assistance of the AI in the second and third rounds. In the fourth round, doctor B and doctor D made diagnoses with the help of the AI and the other two doctors without the help of the AI. In the last round, doctor A and doctor B made diagnoses with the help of AI and the other two doctors without the help of the AI. Results. Doctor A, doctor B, and doctor D had a higher accuracy in the diagnosis of glaucoma with the assistance of AI in the second ( p = 0.036 , p = 0.003 , and p ≤ 0.000 ) and the third round ( p = 0.021 , p ≤ 0.000 , and p ≤ 0.000 ) than in the first round. The accuracy of at least one doctor was higher than that of AI in the second and third rounds, in spite of no detectable significance ( p = 0.283 , p = 0.727 , p = 0.344 , and p = 0.508 ). The four doctors’ overall accuracy ( p = 0.004 and p ≤ 0.000 ) and sensitivity ( p = 0.006 and p ≤ 0.000 ) as a whole were significantly improved in the second and third rounds. Conclusions. This “Doctor + AI” model can clarify the role of doctors and AI in medical responsibility and ensure the safety of patients, and importantly, this model shows great potential and application prospects.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Practical Application of Artificial Intelligence Technology in Glaucoma Diagnosis

Gong

Yin

et al. 2022

Journal of Ophthalmology

Self Cite

View full text Add to dashboard Cite

show abstract

“…In [105], the authors developed a hierarchical deep learning system (HDLS) using 1791 fundus photographs for glaucoma diagnosis. Its recognition accuracy was 53% for the optic cup, 12% for the optic disc, and 16% for retinal nerve fiber layer defects.…”

Section: Segmentation Of the Outer Limits Of The Optic Disc And The Excavation By Detecting Glaucomatous Features Of The Papillamentioning

confidence: 99%

Literature Review on Artificial Intelligence Methods for Glaucoma Screening, Segmentation, and Classification

et al. 2022

View full text Add to dashboard Cite

Artificial intelligence techniques are now being applied in different medical solutions ranging from disease screening to activity recognition and computer-aided diagnosis. The combination of computer science methods and medical knowledge facilitates and improves the accuracy of the different processes and tools. Inspired by these advances, this paper performs a literature review focused on state-of-the-art glaucoma screening, segmentation, and classification based on images of the papilla and excavation using deep learning techniques. These techniques have been shown to have high sensitivity and specificity in glaucoma screening based on papilla and excavation images. The automatic segmentation of the contours of the optic disc and the excavation then allows the identification and assessment of the glaucomatous disease’s progression. As a result, we verified whether deep learning techniques may be helpful in performing accurate and low-cost measurements related to glaucoma, which may promote patient empowerment and help medical doctors better monitor patients.

show abstract

“…DL has been often applied in the detection of glaucoma. There have been many reports of the use of AI in the diagnosis via structural changes, including retinal fundus photos ( 6 – 11 ) and optical coherence tomography (OCT) ( 12 – 14 ). Previous studies have focused on function changes as well as VF ( 15 – 18 ) loss.…”

Section: Introductionmentioning

confidence: 99%

A Structure-Related Fine-Grained Deep Learning System With Diversity Data for Universal Glaucoma Visual Field Grading

et al. 2022

View full text Add to dashboard Cite

PurposeGlaucoma is the main cause of irreversible blindness worldwide. However, the diagnosis and treatment of glaucoma remain difficult because of the lack of an effective glaucoma grading measure. In this study, we aimed to propose an artificial intelligence system to provide adequate assessment of glaucoma patients.MethodsA total of 16,356 visual fields (VFs) measured by Octopus perimeters and Humphrey Field Analyzer (HFA) were collected, from three hospitals in China and the public Harvard database. We developed a fine-grained grading deep learning system, named FGGDL, to evaluate the VF loss, compared to ophthalmologists. Subsequently, we discuss the relationship between structural and functional damage for the comprehensive evaluation of glaucoma level. In addition, we developed an interactive interface and performed a cross-validation study to test its auxiliary ability. The performance was valued by F1 score, overall accuracy and area under the curve (AUC).ResultsThe FGGDL achieved a high accuracy of 85 and 90%, and AUC of 0.93 and 0.90 for HFA and Octopus data, respectively. It was significantly superior (p < 0.01) to that of medical students and nearly equal (p = 0.614) to that of ophthalmic clinicians. For the cross-validation study, the diagnosis accuracy was almost improved (p < 0.05).ConclusionWe proposed a deep learning system to grade VF of glaucoma with a high detection accuracy, for effective and adequate assessment for glaucoma patients. Besides, with the convenient and credible interface, this system can promote telemedicine and be used as a self-assessment tool for patients with long-duration diseases.

show abstract

A hierarchical deep learning approach with transparency and interpretability based on small samples for glaucoma diagnosis

Cited by 28 publications

References 29 publications

Practical Application of Artificial Intelligence Technology in Glaucoma Diagnosis

Practical Application of Artificial Intelligence Technology in Glaucoma Diagnosis

Literature Review on Artificial Intelligence Methods for Glaucoma Screening, Segmentation, and Classification

A Structure-Related Fine-Grained Deep Learning System With Diversity Data for Universal Glaucoma Visual Field Grading

Contact Info

Product

Resources

About