Characterization of Central Visual Field Loss in End-stage Glaucoma by Unsupervised Artificial Intelligence

Wang, Mengyu; Tichelaar, Jorryt; Pasquale, Louis R.; Shen, Lucy Q.; Boland, Michael V.; Wellik, Sarah R.; Moraes, Carlos Gustavo De; Myers, Jonathan S.; Ramulu, Pradeep Y.; Kwon, MiYoung; Saeedi, Osamah; Wang, Hui; Baniasadi, Neda; Li, Dian; Bex, Peter J.; Elze, Tobias

doi:10.1001/jamaophthalmol.2019.5413

Cited by 45 publications

(25 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Each total deviation (TD) plot was decomposed into 16 archetypal patterns including one normal VF pattern and 15 defect patterns determined in our prior work, 42 which were clinically validated in a subsequent study 45 and further applied to improve glaucoma diagnosis and progression detection. 39,[46][47][48] An illustration of the 16 archetype (AT) patterns and corre-sponding nomenclature can be found in Figure 1A. An example of VF decomposition into ATs is shown in Figure 1B.…”

Section: Statistical Analysesmentioning

confidence: 99%

Inter-Eye Association of Visual Field Defects in Glaucoma and Its Clinical Utility

Teng

Choi

et al. 2020

Trans. Vis. Sci. Tech.

Self Cite

View full text Add to dashboard Cite

Section: Statistical Analysesmentioning

confidence: 99%

Inter-Eye Association of Visual Field Defects in Glaucoma and Its Clinical Utility

Teng

Choi

et al. 2020

Trans. Vis. Sci. Tech.

Self Cite

View full text Add to dashboard Cite

“…Furthermore, artificial intelligence tools like convolutional neural networks and other deep-learning approaches have been applied to study the onset and progression of glaucoma, both structurally and functionally. 54 – 56 Our results suggest that substantial information about the disease onset is available before the development of functional deficits. As deep-learning approaches are specialized in exploiting such “hidden” information to classify and predict outcomes, our findings also encourage further deep-learning applications to predict the onset of future functional vision loss from retinal structure.…”

Section: Discussionmentioning

confidence: 73%

Three-dimensional Neuroretinal Rim Thickness and Visual Fields in Glaucoma: A Broken-stick Model

Liu

McClurkin

Tsikata

et al. 2020

Journal of Glaucoma

Self Cite

View full text Add to dashboard Cite

I. Analysis of the structure-function relationship in the linear scale with the broken-stick model We also applied the broken-stick model to the VF data and the MDB structural parameters both in the linear scale. When the VF TD data was transformed to the linear scale, the observation that "the structurefunction relationship displayed a plateau, with neuroretinal rim thickness values being unrelated to VF values" was not readily visualized in some of quadrants/ sectors (Figure S1). Further statistical evaluation shows that the slopes above the tipping point are not significantly different from 0, as shown in Table S1. We added in a model-agnostic lowess smoothing curve to compare with the fitted broken-stick model on the data of unlogged VF TD against the MDB. The broken stick model fit didn't match with the lowess fit for the Temporal-Inferior region and the Inferior quadrant. Furthermore, the breakpoint didn't exist for the nasal and temporal quadrants and the TS and NI octants, as shown in the table below where the pvalues are greater than 0.05 (Table S1). For the breakpoints which did exist (i.e. the global value, superior and inferior quadrants, TI and NS sectors), the fitted breakpoints in the linear scale were different for the breakpoints obtained in the logarithmic fits and not encompassed by their respective confidence intervals from the logarithmic scale. As a final remark, indeed the question of whether the structure-function relationship between the VF data and the neuroretinal rim data should be established as a logarithmic-linear scale or as a linear-linear scale has been actively studied in the past two decades. 1-4 A variety of models have been employed to describe the structure-function relationship in both the linear and curvilinear regimes. With this rich history of research on this topic behind us, we primarily presented the data by looking at the VF in the logarithmic scale (or in its native "untransformed" form) since it is more clinically utilized. Also, the logarithmic scale helps to distinguish among the moderate and severe glaucomatous patients.

show abstract

“…This approach has previously been used to classify fields in glaucoma, as well as to detect progressive change over time. [17][18][19][32][33][34][35][36] Finally, semisupervised learning uses a combination of the two approaches, where one has a relatively small set of labeled data and generally a much larger amount of unlabeled data. The labeled dataset is used to obtain a reasonable initial model, which is then used to perform predictions on the unlabeled dataset.…”

Section: Training Validation and Testing Of Deep Learning Modelsmentioning

confidence: 99%

“…The authors showed that the patterns detected by their technique, such as arcuate, partial arcuate, etc., corresponded well to classification by human graders in the Ocular Hypertension Treatment Study. In a follow-up study, Wang et al 32 proposed to use archetypal analysis to classify central visual field patterns in glaucoma. It should be noted, however, that archetypal analysis is a statistical technique closely resembling traditional factor analysis and bearing no relationship to deep learning artificial neural networks.…”

Section: Introductionmentioning

confidence: 99%

A Review of Deep Learning for Screening, Diagnosis, and Detection of Glaucoma Progression

Thompson

Jammal

Medeiros

2020

Trans. Vis. Sci. Tech.

130

View full text Add to dashboard Cite

Because of recent advances in computing technology and the availability of large datasets, deep learning has risen to the forefront of artificial intelligence, with performances that often equal, or sometimes even exceed, those of human subjects on a variety of tasks, especially those related to image classification and pattern recognition. As one of the medical fields that is highly dependent on ancillary imaging tests, ophthalmology has been in a prime position to witness the application of deep learning algorithms that can help analyze the vast amount of data coming from those tests. In particular, glaucoma stands as one of the conditions where application of deep learning algorithms could potentially lead to better use of the vast amount of information coming from structural and functional tests evaluating the optic nerve and macula. The purpose of this article is to critically review recent applications of deep learning models in glaucoma, discussing their advantages but also focusing on the challenges inherent to the development of such models for screening, diagnosis and detection of progression. After a brief general overview of deep learning and how it compares to traditional machine learning classifiers, we discuss issues related to the training and validation of deep learning models and how they specifically apply to glaucoma. We then discuss specific scenarios where deep learning has been proposed for use in glaucoma, such as screening with fundus photography, and diagnosis and detection of glaucoma progression with optical coherence tomography and standard automated perimetry. Translational Relevance Deep learning algorithms have the potential to significantly improve diagnostic capabilities in glaucoma, but their application in clinical practice requires careful validation, with consideration of the target population, the reference standards used to build the models, and potential sources of bias.

show abstract

Characterization of Central Visual Field Loss in End-stage Glaucoma by Unsupervised Artificial Intelligence

Cited by 45 publications

References 34 publications

Inter-Eye Association of Visual Field Defects in Glaucoma and Its Clinical Utility

Inter-Eye Association of Visual Field Defects in Glaucoma and Its Clinical Utility

Three-dimensional Neuroretinal Rim Thickness and Visual Fields in Glaucoma: A Broken-stick Model

A Review of Deep Learning for Screening, Diagnosis, and Detection of Glaucoma Progression

Contact Info

Product

Resources

About