Using unsupervised learning with variational bayesian mixture of factor analysis to identify patterns of glaucomatous visual field defects.

PURPOSE To determine whether an unsupervised machine learning classifier can identify patterns of visual field loss in standard visual fields consistent with typical patterns learned by decades of human experience. METHODS Standard perimetry thresholds for 52 locations plus age from one eye of each of 156 patients with glaucomatous optic neuropathy (GON) and 189 eyes of healthy subjects were clustered with an unsupervised machine classifier, variational Bayesian mixture of factor analysis (vbMFA). RESULTS The vbMFA formed five distinct clusters. Cluster 5 held 186 of 189 fields from normal eyes plus 46 from eyes with GON. These fields were then judged within normal limits by several traditional methods. Each of the other four clusters could be described by the pattern of loss found within it. Cluster 1 (71 GON + 3 normal optic discs) included early, localized defects. A purely diffuse component was rare. Cluster 2 (26 GON) exhibited primarily deep superior hemifield defects, and cluster 3 (10 GON) held deep inferior hemifield defects only or in combination with lesser superior field defects. Cluster 4 (6 GON) showed deep defects in both hemifields. In other words, visual fields within a given cluster had similar patterns of loss that differed from the predominant pattern found in other clusters. The classifier separated the data based solely on the patterns of loss within the fields, without being guided by the diagnosis, placing 98.4% of the healthy eyes within the same cluster and spreading 70.5% of the eyes with GON across the other four clusters, in good agreement with a glaucoma expert and pattern standard deviation. CONCLUSIONS Without training-based diagnosis (unsupervised learning), the vbMFA identified four important patterns of field loss in eyes with GON in a manner consistent with years of clinical experience.

[1]  Terrence J. Sejnowski,et al.  Variational Learning of Clusters of Undercomplete Nonsymmetric Independent Components , 2003, J. Mach. Learn. Res..

[2]  A Heijl,et al.  Glaucoma Hemifield Test. Automated visual field evaluation. , 1992, Archives of ophthalmology.

[3]  A E Maumenee,et al.  Optic disc parameters and onset of glaucomatous field loss. I. Methods and progressive changes in disc morphology. , 1979, Archives of ophthalmology.

[4]  Chris A. Johnson,et al.  The Relationship Between Structural and Functional Alterations in Glaucoma: A Review , 2000, Seminars in ophthalmology.

[5]  C. Johnson,et al.  Visual field damage in normal-tension and high-tension glaucoma. , 1989, American journal of ophthalmology.

[6]  W. Green,et al.  Optic nerve damage in human glaucoma. III. Quantitative correlation of nerve fiber loss and visual field defect in glaucoma, ischemic neuropathy, papilledema, and toxic neuropathy. , 1982, Archives of ophthalmology.

[7]  D B Henson,et al.  Frequency Distribution of Early Glaucomatous Visual Field Defects , 1986, American journal of optometry and physiological optics.

[8]  Shigeo Abe DrEng.  Pattern Classification , 2001, Springer London.

[9]  David G. Stork,et al.  Pattern Classification , 1973 .

[10]  J. Caprioli,et al.  Patterns of early visual field loss in open-angle glaucoma. , 1986, Transactions of the American Ophthalmological Society.

[11]  Chris A. Johnson,et al.  Structure and function evaluation (SAFE): II. Comparison of optic disk and visual field characteristics. , 2003, American journal of ophthalmology.

[12]  D. R. Anderson,et al.  Early foveal involvement and generalized depression of the visual field in glaucoma. , 1984, Archives of ophthalmology.

[13]  A Heijl,et al.  Lack of diffuse loss of differential light sensitivity in early glaucoma , 1989, Acta ophthalmologica.

[14]  P A Sample,et al.  Visual function-specific perimetry for indirect comparison of different ganglion cell populations in glaucoma. , 2000, Investigative ophthalmology & visual science.

[15]  S. Drance,et al.  The disc and the field in glaucoma. , 1978, Ophthalmology.

[16]  J. Flammer,et al.  Is there a difference between glaucoma patients with rather localized visual field damage and patients with more diffuse visual field damage , 1987 .

[17]  S M Drance,et al.  The early field defects in glaucoma. , 1969, Investigative ophthalmology.

[18]  Morin Jd,et al.  Changes in the visual fields in glaucoma: static and kinetic perimetry in 2,000 patients. , 1979 .

[19]  Robert N Weinreb,et al.  Comparing machine learning classifiers for diagnosing glaucoma from standard automated perimetry. , 2002, Investigative ophthalmology & visual science.

[20]  B. Chauhan,et al.  Diffuse and localized glaucomatous field loss in light-sense, flicker and resolution perimetry , 2004, Graefe's Archive for Clinical and Experimental Ophthalmology.

[21]  B. Chauhan,et al.  Diffuse loss of sensitivity in early glaucoma. , 1999, Investigative ophthalmology & visual science.

[22]  Chris A. Johnson,et al.  The Ocular Hypertension Treatment Study: baseline factors that predict the onset of primary open-angle glaucoma. , 2002, Archives of ophthalmology.

[23]  Robert N Weinreb,et al.  Using machine learning classifiers to identify glaucomatous change earlier in standard visual fields. , 2002, Investigative ophthalmology & visual science.

[24]  S M Drance,et al.  The glaucomatous visual field. , 1972, Investigative ophthalmology.

[25]  Werner Eb,et al.  Location of early glaucomatous visual field defects. , 1980 .

[26]  R. Susanna,et al.  Use of Discriminant Analysis to Identify Unknown Author , 1972 .

[27]  Douglas R. Anderson.  Automated Static Perimetry , 1992 .

[28]  Michael A. Arbib,et al.  The handbook of brain theory and neural networks , 1995, A Bradford book.

[29]  David G. Stork,et al.  Pattern Classification (2nd ed.) , 1999 .

[30]  R. Leblanc,et al.  Repeatable diffuse visual field loss in open-angle glaucoma. , 1997, Ophthalmology.

[31]  Douglas R. Anderson,et al.  Automatic Perimetry in Glaucoma: A Practical Guide , 1985 .

[32]  R. Weinreb,et al.  The structure-function relationship in eyes with glaucomatous visual field loss that crosses the horizontal meridian. , 2002, Archives of ophthalmology.

[33]  Zoubin Ghahramani,et al.  Variational Inference for Bayesian Mixtures of Factor Analysers , 1999, NIPS.

[34]  B. Becker,et al.  The onset and evolution of glaucomatous visual field defects. , 1982, Ophthalmology.

[35]  A. Heijl,et al.  THE FREQUENCY DISTRIBUTION OF EARLIEST GLAUCOMATOUS VISUAL FIELD DEFECTS DOCUMENTED BY AUTOMATIC PERIMETRY , 1984, Acta ophthalmologica.

[36]  Michael I. Jordan,et al.  An Introduction to Variational Methods for Graphical Models , 1999, Machine-mediated learning.

[37]  David Mackay,et al.  Probable networks and plausible predictions - a review of practical Bayesian methods for supervised neural networks , 1995 .

[38]  M F Armaly.  Visual field defects in early open angle glaucoma. , 1971, Transactions of the American Ophthalmological Society.

[39]  S M Drance,et al.  Early visual field disturbances in glaucoma. , 1977, Archives of ophthalmology.

[40]  Terrence J. Sejnowski,et al.  Comparison of machine learning and traditional classifiers in glaucoma diagnosis , 2002, IEEE Transactions on Biomedical Engineering.

[41]  J. Morin.  Changes in the visual fields in glaucoma: static and kinetic perimetry in 2,000 patients. , 1979, Transactions of the American Ophthalmological Society.