Split Bundle Detection in Polarimetric Images of the Human Retinal Nerve Fiber Layer

Summary

Objectives: One method for assessing pathological retinal nerve fiber layer (NFL) appearance is by comparing the NFL to normative values, derived from healthy subjects. These normative values will be more specific when normal physiological differences are taken into account. One common variation is a split bundle. This paper describes a method to automatically detect these split bundles.

Methods: The thickness profile along the NFL bundle is described by a non-split and a split bundle model. Based on these two fits, statistics are derived and used as features for two non-parametric classifiers (Parzen density based and k nearest neighbor). Features were selected by forward feature selection. Three hundred and nine superior and 324 inferior bundles were used to train and test this method.

Results: The prevalence of split superior bundles was 68% and the split inferior bundles’ prevalence was 13%. The resulting estimated error of the Parzen density-based classifier was 12.5% for the superior bundle and 10.2% for the inferior bundle. The k nearest neighbor classifier errors were 11.7% and 9.2%.

Conclusions: The classification error of automated detection of split inferior bundles is not much smaller than its prevalence, thereby limiting the usefulness of separate cut-offvalues for split and non-split inferior bundles. For superior bundles, however, the classification error was low compared to the prevalence. Application of specific cut-offvalues, selected by the proposed classification system, may therefore increase the specificity and sensitivity of pathological NFL detection.

Keywords

Pattern recognition system - computer-assisted diagnosis - glaucoma - retina - morphology

• References

• 1 Dreher AW, Reiter K. Scanning laser polarimetry of the retinal nerve fiber layer. Polarization Analysis and Measurement. Vol. 1746 of Proc. SPIE 1992: 34-41.
• 2 Zhou Q, Knighton RW. Light scattering and form birefringence of parallel cylindrical arrays that represent cellular organelles of the retinal nerve fiber layer. Applied Optics 1997; 36 (10) 2273-85.
• 3 Colen TP, Lemij HG. Prevalence of split nerve fiber layer bundles in healthy eyes imaged with scanning laserpolarimetry. Ophthalmology 2001; 108 (01) 151-6.
• 4 Peters A, Lausen B, Michelson G, Gefeller O. Diagnosis of glaucoma by indirect classifiers. Methods Inf Med 2003; 42 (01) 99-103.
• 5 Adler W, Hothorn T, Lausen B. Simulation based analysis of automated classification of medical images. Methods Inf Med 2004; 43 (02) 150-5.
• 6 Chrastek R, Skokan M, Kubecka L, Wolf M, Donath K, Jan J. et al. Multimodal Retinal Image Registration for Optic Disk Segmentation. Methods Inf Med 2004; 43 (04) 336-42.
• 7 Vermeer KA, Vos FM, Lemij HG, Vossepoel AM. A model based method for retinal blood vessel detection. Comput Biol Med 2004; 34 (03) 209-19.
• 8 Knighton RW, Huang XR, Greenfield DS. Analytical Model of Scanning Laser Polarimetry for Retinal Nerve Fiber Layer Assessment. Invest Ophthalmol Vis Sci 2002; 43: 383-92.
• 9 Parzen E. On the estimation of a probability density function and mode. Ann Math Stat 1962; 33: 1065-76.
• 10 Müller HG. Ihm Ε Kernel estimation techniques for the analysis of clinical curves. Methods Inf Med 1985; 24 (04) 218-24.
• 11 Webb AR. Ch. 3. Statistical Pattern Recognition. 2. Chichester, UK: Wiley; 2002.
• 12 Vapnik VN. Statistical Learning Theory. John Wiley & Sons, Inc.; 1998.