Research Article| Volume 43, ISSUE 4, P366-372, August 2020

Download started.


Robust keratoconus detection with Bayesian network classifier for Placido-based corneal indices

Published:December 19, 2019DOI:


      • A naïve Bayes classifier for keratoconus detection is presented.
      • It uses primary Placido-based corneal indices, described in literature and computed directly from the image of the disks reflected on the cornea.
      • The classifier has been tested and compared with deterministic indices on a population of 60 corneas, as well as under artificially noisy data.
      • It showed excellent discrimination capability and robustness.
      • The developed approach is simple, platform-independent and reproducible on any Placido-based topographer.



      To evaluate in a sample of normal and keratoconic eyes a simple Bayesian network classifier for keratoconus identification that uses previously developed topographic indices, calculated directly from the digital analysis of the Placido ring images.


      A comparative study was performed on a total of 60 eyes from 60 patients (age 20–60 years) from the Department of keratoconus of INVISION Ophthalmology clinic (Almería, Spain). Patients were divided into two groups depending on their preliminary diagnosis based on the classical topographic criteria: a control group without topographic alteration (30 eyes) and a keratoconus group (30 eyes). The keratoconus group included all grades except grade IV with excessively distorted corneal topography. All cases were examined using the CSO topography system (CSO, Firenze, Italy), and primary corneal Placido-indices were computed, as described in literature. Finally, a classifier was built by fitting a conditional linear Gaussian Bayesian network to the data, using the 5- and 10-fold cross-validation. For comparison, the original data were perturbed with random white noise of different magnitude.


      The naïve Bayes classifier showed perfect discrimination ability among normal and keratoconic corneas, with 100% of sensibility and specificity, even in the presence of a very significant noise.


      The Bayesian network classifiers are highly accurate and proved a stable screening method to assist ophthalmologists with the detection of keratoconus, even in the presence of noise or incomplete data. This algorithm is easily implemented for any Placido topographic system.


      To read this article in full you will need to make a payment

      Purchase one-time access:

      Academic & Personal: 24 hour online accessCorporate R&D Professionals: 24 hour online access
      One-time access price info
      • For academic or personal research use, select 'Academic and Personal'
      • For corporate R&D use, select 'Corporate R&D Professionals'


      Subscribe to Contact Lens and Anterior Eye
      Already a print subscriber? Claim online access
      Already an online subscriber? Sign in
      Institutional Access: Sign in to ScienceDirect


        • Mejía-Barbosa Y.
        • Malacara-Hernández D.
        A review of methods for measuring corneal topography.
        Optomet Vision Sci. 2001; 78: 240-253
        • van Saarlos P.P.
        • Constable I.J.
        Improved method for calculation of corneal topography for any photokeratoscope geometry.
        Optomet Vision Sci. 1991; 68: 960-965
        • Klein S.A.
        A corneal topography algorithm that produces continuous curvature.
        Optomet Vision Sci. 1992; 69: 829-834
        • Halstead M.A.
        • Barsky B.
        • Klein S.A.
        • Mandell R.B.
        A spline surface algorithm for reconstruction of corneal topography from a videokeratographic reflection pattern.
        Optomet Vision Sci. 1995; 72: 821-827
        • Klein S.A.
        Corneal topography reconstruction algorithm that avoids the skew ray ambiguity and the skew ray error.
        Optomet Vision Sci. 1997; 74: 945-962
        • Sicam V.A.D.P.
        • Van der Heijde R.G.L.
        Topographer reconstruction of the nonrotation-symmetric anterior corneal surface features.
        Optom Vis Sci. 2006; 83: 910-918
        • Rabinowitz Y.S.
        Surv Ophthalmol. 1998; 42: 297-319
        • Piñero D.
        • Alió J.L.
        • Barraquer R.I.
        • Michael R.
        • Jiménez R.
        Corneal biomechanics, refraction, and corneal aberrometry in keratoconus: an integrated study.
        Invest Ophthalmol Vis Sci. 2010; 51: 1948-1955
        • Shah S.
        • Laiquzzaman M.
        • Bhojwani R.
        • Mantry S.
        • Cunliffe I.
        Assessment of the biomechanical properties of the cornea with the Ocular Response Analyzer in normal and keratoconic eyes.
        Invest Ophthalmol Vis Sci. 2007; 48: 3026-3031
        • Meek K.
        • Tuft S.
        • Huang Y.
        • Gill P.
        • Hayes S.
        • Newton R.
        • Bron A.
        Changes in collagen orientation and distribution in keratoconus corneas.
        Invest Ophthalmol Vis Sci. 2005; 46: 1948-1956
        • Daxer A.
        • Fratzl P.
        Collagen fibril orientation in the human corneal stroma and its implication in keratoconus.
        Invest Ophthalmol Vis Sci. 1997; 38: 121-129
        • Daxer A.
        • Fratzl P.
        Erratum: Collagen fibril orientation in the human corneal stroma and its implication in keratoconus (investigate ophthalmology and visual science (1997) 38 (121-129)).
        Invest Ophthalmol Vis Sci. 1997; 38
        • Smolek M.
        • Beekhuis W.
        Collagen fibril orientation in the human corneal stroma and its implications in keratoconus.
        Invest Ophthalmol Vis Sci. 1997; 38: 1289-1290
        • Alió J.L.
        • Shabayek M.H.
        Corneal higher order aberrations: a method to grade keratoconus.
        J Refract Surg. 2006; 22: 539-545
        • Gobbe M.
        • Guillon M.
        Corneal wavefront aberration measurements to detect keratoconus patients.
        Cont Lens Anterior Eye. 2005; 28: 57-66
        • Barbero S.
        • Marcos S.
        • Merayo-Lloves J.
        • Moreno-Barriuso E.
        Validation of the estimation of corneal aberrations from videokeratography in keratoconus.
        J Refract Surg. 2002; 18: 263-270
        • Ambrósio R.
        • Lopes B.
        • Faria-Correia F.
        • Salomão M.
        • Bühren J.
        • Roberts C.
        • Elsheikh A.
        • Vinciguerra R.
        • Vinciguerra P.
        Integration of Scheimpflug-based corneal tomography and biomechanical assessments for enhancing ectasia detection.
        J Refract Surg. 2017; 33: 434-443
        • Shetty R.
        • Rao H.
        • Khamar P.
        • Sainani K.
        • Vunnava K.
        • Jayadev C.
        • Kaweri L.
        Keratoconus screening indices and their diagnostic ability to distinguish normal from ectatic corneas.
        Am J Ophthalmol. 2017; 181: 140-148
        • Steinberg J.
        • Amirabadi N.
        • Frings A.
        • Mehlan J.
        • Katz T.
        • Linke S.
        Keratoconus screening with dynamic biomechanical in vivo Scheimpflug analyses: A proof-of-concept study.
        J Refract Surg. 2017; 33: 773-778
        • Martínez-Abad A.
        • Piñero D.
        New perspectives on the detection and progression of keratoconus.
        J Cataract Refract Surg. 2017; 43: 1213-1227
        • Huseynli S.
        • Salgado-Borges J.
        • Alió J.L.
        Comparative evaluation of Scheimpflug tomography parameters between thin non-keratoconic, subclinical keratoconic, and mild keratoconic corneas.
        Eur J Ophthalmol. 2018; 28: 521-534
        • Piñero D.
        • Alió J.
        • Alesón A.
        • Escaf M.
        • Miranda M.
        Pentacam posterior and anterior corneal aberrations in normal and keratoconic eyes.
        Clin Exp Optomet. 2009; 92: 297-303
        • Carvalho L.A.
        Preliminary results of neural networks and Zernike polynomials for classification of videokeratography maps.
        Optomet Vision Sci. 2005; 82: 151-158
        • Accardo P.A.
        • Pensiero S.
        Neural network-based system for early keratoconus detection from corneal topography.
        J Biomed Inform. 2002; 35: 151-159
        • Issarti I.
        • Consejo A.
        • Jiménez-García M.
        • Hershko S.
        • Koppen C.
        • Rozema J.
        Computer aided diagnosis for suspect keratoconus detection.
        Comput Biol Med. 2019; 109: 33-42
        • Ramos-López D.
        • Martínez-Finkelshtein A.
        • Castro-Luna G.M.
        • Piñero D.
        • Alió J.L.
        Placido-based indices of corneal irregularity.
        Optomet Vision Sci. 2011; 88: 1220-1231
        • Ramos-López D.
        • Martínez-Finkelshtein A.
        • Castro-Luna G.M.
        • Burgera-Giménez N.
        • Vega-Estrada A.
        • Piñero D.
        • Alió J.L.
        Screening subclinical keratoconus with Placido-based corneal indices.
        Optomet Vision Sci. 2013; 90: 335-343
        • Larrañaga P.
        • Moral S.
        Probabilistic graphical models in artificial intelligence.
        Appl Soft Comput. 2011; 11: 1511-1528
        • Friedman N.
        • Geiger D.
        • Goldszmidt M.
        Bayesian network classifiers.
        Mach Learn. 1997; 29: 131-163
        • Bressan G.N.
        • Oliveira V.A.
        • Hruschka E.R.
        • Nicoletti M.C.
        Using Bayesian networks with rule extraction to infer risk of weed infestation in a corn-crop.
        Eng Appl Artif Intel. 2009; 22: 579-592
        • Markus M.
        • Hejazi I.
        • Bajcsy P.
        • Giustolisi O.
        • Savic A.
        Prediction of weekly nitrate-N fluctuations in a small agricultural watershed in Illinois.
        J Hydroinformat 12.3. 2010; : 251-261
        • Fytilis N.
        • Rizzo D.M.
        Coupling self-organizing maps with a Naïve Bayesian classifier: Stream classification studies using multiple assessment data.
        Water Resour Res. 2013; 49: 7747-7762
        • Maldonado A.D.
        • Uusitalo L.
        • Tucker A.
        • Blenckner T.
        • Aguilera P.A.
        • Salmerón A.
        Prediction of a complex system with few data: Evaluation of the effect of model structure and amount of data with dynamic bayesian network models.
        Environ Model Softw. 2019; 118: 281-297
        • Pearl J.
        Probabilistic reasoning in intelligent systems.
        Morgan-Kaufmann, 1988
        • Fung R.
        • Chang K.C.
        Weighting and integrating evidence for stochastic simulation in Bayesian networks.
        Uncertainty Artif Intel. 1990; 5: 209-220
        • Stone M.
        Cross-validatory choice and assessment of statistical predictions.
        J Royal Stat Soci. Series B (Methodological). 1974; 36: 111-147
        • Scutari M.
        Learning bayesian networks with the bnlearn R package.
        J Stat Softw. 2010; 35: 1-22
        • Ramos-López D.
        • Maldonado A.D.
        Placido analysis of corneal irregularity, R package version 0.1.1, Tech. rep..
        University of Almeria, 2019