Diagnosis of Glaucoma by Indirect Classifiers

Summary

Objectives: Demonstration of the applicability of a framework called indirect classification to the example of glaucoma classification. Indirect classification combines medical a priori knowledge and statistical classification methods. The method is compared to direct classification approaches with respect to the estimated misclassification error.

Methods: Indirect classification is applied using classification trees and the diagnosis of glaucoma. Misclassification errors are reduced by bootstrap aggregation. As direct classification methods linear discriminant analysis, classification trees and bootstrap aggregated classification trees are utilized in the problem of glaucoma diagnosis. Misclassification rates are estimated via 10-fold cross-validation.

Results: Indirect classification techniques reduce the misclassification error in the context of glaucoma classification compared to direct classification methods.

Conclusions: Embedding a priori knowledge into statistical classification techniques can improve misclassification results. Indirect classification offers a framework to realize this combination.

Keywords

Glaucoma - classification - decision trees

• References

• 1 Hand DJ, Li HG, Adams NM. Supervised classification with structured class definitions. Computational Statistics & Data Analysis 2001; 36: 209-25.
• 2 Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and regression trees. California: Wadsworth; 1984
• 3 Breiman L. Bagging predictors. Machine learning 1996; 24 (Suppl. 02) 123-40.
• 4 Breiman L. Arcing classifiers. The Annals of Statistics 1998; 26 (Suppl. 03) 801-49.
• 5 Coleman AL. Glaucoma. The Lancet 1999; 354: 1803-10.
• 6 Weih LM, Nanjan M, McCarty CA, Taylor HR. Prevalence and predictors of open-angle glaucoma : Results from the visual impairment project. Ophthalmology 2001; 108 (Suppl. 11) 1966-72.
• 7 Lee BL, Bathija R, Weinreb RN. The definition of normal-tension glaucoma. J Glaucoma 1998; 7 (Suppl. 06) 366-71.
• 8 Horn FK, Jonas JB, Korth M, Jünemann A, Grundler A. The full-field flicker test in early diagnosis of chronic open-angle glaucoma. Am J Ophthalmology 1997; 123 (Suppl. 03) 313-9.
• 9 Mardin CY, Horn FK, Jonas JB, Budde WM. Preperimetric glaucoma diagnosis by confocal scanning laser tomography of the optic disc. Br J Ophthalmology 1999; 83 (Suppl. 03) 299-304.
• 10 Swindale NV, Stjepanovic G, Chin A, Mikelberg FS. Automated analysis of normal and glaucomatous optic nerve head topography images. Investigative Ophthalmology Visual Science 2000; 41 (Suppl. 07) 1730-42.
• 11 Heidelberg-Engineering. Heidelberg Retina Tomograph: Bedienungsanleitung Software version 2.01. Heidelberg: Heidelberg Engineering GmbH; 1997
• 12 Freund Y, Schapire R. Experiments with a new boosting algorithm. Machine Learning Proceedings of the Thirteenth International Conference.; 1996: 148-56.
• 13 Schiavo RA, Hand DJ. Ten more years of error rate research. International Statistical Review 2000; 68 (Suppl. 03) 295-310.
• 14 Therneau TM, Atkinson EJ. An introduction to recursive partitioning using the rpart routine. Technical Report 61. Rochester: Section of Biostatistics, Mayo Clinic; 1997
• 15 Ihaka R, Gentleman R. R:A Language for data analysis and graphics. J Computational Graphical Statistics 1996; 5: 299-314.
• 16 Eid TM, Spaeth GL, Katz LJ, Azuara-Blanco A, Agusburger J, Nicholl J. Quantitative estimation of retinal nerve fiber layer height in glaucoma and the relationship with optic nerve head topography and visual field. J Glaucoma 1997; 6 (Suppl. 04) 221-30.
• 17 Iester M, Mikelberg FS, Courtright P, Drance SM. Correlation between the visual field indices and heidelberg retina tomograph parameters. J Glaucoma 1997; 6 (Suppl. 02) 78-82.
• 18 Kamal DS, Viswanathan AC, Garway-Heath DF, Hitchings RA, Poinoosawmy D, Bunce C. Detection of optic disc change with the Heidelberg retina tomograph before confirmed visual field change in ocular hypertensives converting to early Glaucoma. Br J Ophthalmology 1999; 83 (Suppl. 03) 290-4.
• 19 Hammond P, Hutton TJ, Nelson-Moon ZL, Hunt NP, Madgwick AJA. Classifying vertical facial deformity using supervised and unsupervised learning. Methods Inf Med 2001; 40: 365-72.
• 20 Cox DR, Wermuth N. Multivariate dependencies. Models, analysis and interpretation. London, UK: Chapman & Hall; 1996
• 21 Dusseldorp E, Meulman JJ. Prediction in medicine by integrating regression trees into regression analysis with optimal scaling. Methods Inf Med 2001; 40: 403-9.
• 22 Lausen B, Sauerbrei W, Schumacher M. Classification and regression trees (CART) used for the exploration of prognostic factors measured on different scales. In: Dirschedl P, Ostermann R. editors. Computational Statistics. Heidelberg: Physica-Verlag; 1994: 483-96.
• 23 McQuatt A, Sleeman D, Andrews PJD, Corruble V, Jones PA. Discussing anomalous situations using decision trees: A head injury study. Methods Inf Med 2001; 40: 373-9.
• 24 Hothorn T, Lausen B. Double-Bagging: Combining classifiers by bootstrap aggregation. Pattern Recognition 2002 36 (6): in press.
• 25 Dudoit S, Fridlyand J, Speed TP. Comparison of discrimination methods for the classification of tumors using gene expression data. J American Statistical Association 2002; 97: 77-87.