REPRODUCIBILITY OF NON--MYDRIATIC … · artery y 1 31 31 31 31 2 32 32 32 32 ... Boya Vegue J....
1
This study was designed to examine the stability of basic eye measurements determined by non- mydriatic retinography and their use for personal identification. Means and standard deviations were calculated for the measurements made in each individual (abscissae and ordinates of the macula, papilla centre and first bifurcation of the superior temporal vein, STV) using the set of valid data illustrated in part in Table 1. The criterion used to identify each individual was the matching of all six variables. For each variable, the Thomson test was used to identify or reject a match. According to the corresponding principal component analysis (PCA), interindividual discrimination depends in high proportion on the abscissae and ordinates of the first bifurcation of the STV. For this reason, we then examined the use of bifurcation of the ST artery. REPRODUCIBILITY OF NON REPRODUCIBILITY OF NON-MYDRIATIC RETINOGRAPHIES: MYDRIATIC RETINOGRAPHIES: IMPLICATIONS FOR PERSONAL IDENTIFICATION SYSTEMS IMPLICATIONS FOR PERSONAL IDENTIFICATION SYSTEMS EVER 2010 EVER 2010 [email protected] [email protected] We estimated the means and deviations of the measurements in each individual (abscissae and ordinates of the first ST artery and vein bifurcations) using the set of valid data. An error was thus produced when one of the six variables measured in an individual did not match the individual’s own pattern. The probability of intra- individual error was estimated from the proportion of errors (one or more discrepancies) in the sample considered (using only the complete datasets available). The study sample comprised a set of 864 retinographies obtained in 26 subjects aged 18 to 32 years, in which 4 characteristic anatomical structures were identified. For each anatomical structure, we determined the coordinates x,y as the distances from a reference point at the origin of both axes. This point was the lower left corner of the retinography. The reproducibility of non-mydriatic retinal image measurements was high for the selected points. However, the abscissae and ordinates of the macula showed practically no discriminatory capacity and produced an increase in the probability of intraindividual error. These two variables have thus been eliminated from subsequent study. Our findings point to the blood vessels of the retina as key structures on which to base the development of new identification security devices. Material Sample Table 1: Database used to examine the disriminatory capacity of retinographic measurements for 2 of the 4 structures identified. The figures indicate the number of measurements obtained in each individual. Table 5: Estimated probabilities of intra-individual error. Alpha 0.01 Number of errors 37 Probability of error 0.04282 Using this method an error is produced if any of the 6 variables in an individual do not match the individual’s own pattern. The results obtained for several significance levels are provided below: Alpha 0.01 0.02 0.05 Probability of error 0.0060 0.0633 0.2048 Table 2: Estimated probabilities of the intra-individual error for different levels of significance for the 6 variables examined However, if we only consider the variables abscissa and ordinate of the first bifurcation of the STV, estimated intra-individual errors are: Alpha 0.01 0.02 0.05 Probability of error 0.0030 0.0181 0.0723 Table 3: Estimated probabilities of the intra-individual error for different levels of significance for the variables abscissa and ordinate of the first bifurcation of the STV. Subject Bifurcation vein x Bifurcation vein y Bifurcation artery x Bifurcation artery y 1 31 31 31 31 2 32 32 32 32 3 29 29 29 29 4 41 41 41 41 5 30 30 30 30 6 19 19 19 19 7 39 39 39 39 8 28 28 28 28 9 30 30 30 30 10 35 35 35 35 11 8 8 8 8 12 58 58 58 58 13 28 28 28 28 14 17 17 17 17 15 30 30 30 30 16 30 30 30 30 17 12 12 12 12 18 38 38 38 38 19 31 31 31 31 20 31 31 31 31 21 43 43 43 43 22 30 30 30 30 23 34 34 34 34 24 44 44 44 44 25 44 44 44 44 26 41 41 41 41 Total 864 864 864 864 These results indicate a lower probability of error when the data considered correspond to structures such as vessel bifurcations rather than less easily- defined structures such as the papilla and macula. 1. Boya Vegue J. Atlas de Histología y Organografía microscópica. Médica Panamericana, España, 2002, 393-407. 2. Corcóstegui B. El fondo de ojo en la medicina práctica, Espasa, España, 1983, 29-45. 3. El Danasoury MA. Prospective bilateral study of night glare after laser in situ keratomileusis with single zone and transition zono oblation. J Refr Surg 199814 (5): 512-516, 4. Morgan MW. Visual function changes in eyes of aged patients. Vision and Aging: General and Clinical Perspectives. Ed, A, A, Rosenblom. New York, 1986; 198-200. 5. Putz R., Pabst R. Sobotta, Atlas de Anatomía Humana. Tomo I: Cabeza, Cuello, Miembro superior. Médica Panamericana, Buenos Aires, Madrid, 2006, 362-370. 6. Rodríguez S., Smith-Agreda J.M. Anatomía de los órganos del lenguaje, visión y audición. Médica Panamericana, España, 1999, 250-270. 7. Young B., Heath J.W. Wheater´s Histología Funcional. Elsevier España S.A., 2005, 382-391 Figure 2: Focusing the TRC-NW200 retinographer The retinographer used was the TRC-NW200 by Topcon that does not require pharmacological mydriasis. Using a corrective lens, images can be obtained in ammetropic eyes of up to - 33 diopters in myopes and +40 in hyperopes. To further constrain the estimation of this probability of error, we can assume that the different variables in the individuals behave as independent variables and then by estimating the probability of an interindividual match for each one (probability of error) the probability of the whole set of variables matching can be calculated (interindividual probability of error). The results obtained for this assumption are shown in Table 4. This study was supported by a formal agreement between the Complutense University of Madrid and Alta Eficacia Tecnología SL. MAC X MAC Y BIF X BIF Y PAP X PAP Y TOTAL 6 variables TOTAL 2 variables Alpha Probability of error 0.01 0.875 0.9 0.0875 0.06667 0.47917 0.63333 0.00139 0.0058 0.02 0.85 0.85833 0.07917 0.05833 0.43333 0.5875 0.00086 0.0046 0.05 0.79167 0.7875 0.07083 0.05 0.3375 0.52083 0.00039 0.0035 Tabe 4: Estimated probabilities of the inter-individual error for different levels of significance assuming the independence of variables. The resolution of the image captured by the CCD camera is 3.15 million pixels. The angle captured is 45º for pupils larger than 4.00 mm and 30º for pupil sizes of 3.70 mm. Non valid retinography Valid retinography Pupil diameter <3.70 mm Pupil diameter >3.70 mm Figure 5: Artery and vein patterns of the retina in the left and right eyes Figure 3: Example of the quality of the images obtained Figure 4: Correct position of the subject for image capture. The outer corner of the eye is aligned with a line in the instrument.
Transcript of REPRODUCIBILITY OF NON--MYDRIATIC … · artery y 1 31 31 31 31 2 32 32 32 32 ... Boya Vegue J....
This study was designed to
examine the stability of basic eye
measurements determined by non-
mydriatic retinography and their use
for personal identification.
Means and standard deviations were calculated for the measurements madein each individual (abscissae and ordinates of the macula, papilla centre and first
bifurcation of the superior temporal vein, STV) using the set of valid dataillustrated in part in Table 1. The criterion used to identify each individual was thematching of all six variables. For each variable, the Thomson test was used to
identify or reject a match.
According to the corresponding principal component analysis (PCA),interindividual discrimination depends in high proportion on the abscissae and
ordinates of the first bifurcation of the STV. For this reason, we then examinedthe use of bifurcation of the ST artery.
REPRODUCIBILITY OF NONREPRODUCIBILITY OF NON--MYDRIATIC RETINOGRAPHIES:MYDRIATIC RETINOGRAPHIES:IMPLICATIONS FOR PERSONAL IDENTIFICATION SYSTEMSIMPLICATIONS FOR PERSONAL IDENTIFICATION SYSTEMS
EVER 2010EVER [email protected]@opt.ucm.es
We estimated the means and deviations of the measurements in eachindividual (abscissae and ordinates of the first ST artery and vein bifurcations)
using the set of valid data.
An error was thus produced when one of the six variables measured in anindividual did not match the individual’s own pattern. The probability of intra-individual error was estimated from the proportion of errors (one or morediscrepancies) in the sample considered (using only the complete datasets
available).
The study sample comprised a set of 864retinographies obtained in 26 subjects aged 18 to 32
years, in which 4 characteristic anatomical structureswere identified.
For each anatomical structure, we determined thecoordinates x,y as the distances from a reference pointat the origin of both axes. This point was the lower leftcorner of the retinography.
The reproducibility of non-mydriatic retinal image measurements was highfor the selected points. However, the abscissae and ordinates of the macula
showed practically no discriminatory capacity and produced an increase in theprobability of intraindividual error. These two variables have thus beeneliminated from subsequent study.
Our findings point to the blood vessels of the retina as key structures onwhich to base the development of new identification security devices.
for personal identification.
Material
Sample
Table 1: Database used to examine the disriminatory capacity of retinographic measurements for 2
of the 4 structures identified. The figures indicate the number of measurements obtained in each
individual.
Table 5: Estimated probabilities of intra-individual error.
Alpha 0.01Number of errors 37
Probability of error 0.04282
Using this method an error is produced if any of the 6 variables in anindividual do not match the individual’s own pattern. The results obtained for
several significance levels are provided below:
Alpha 0.01 0.02 0.05
Probability of error 0.0060 0.0633 0.2048Table 2: Estimated probabilities of the intra-individual error for different levels of
significance for the 6 variables examined
However, if we only consider the variables abscissa and ordinate of the firstbifurcation of the STV, estimated intra-individual errors are:
Alpha 0.01 0.02 0.05Probability of error 0.0030 0.0181 0.0723
Table 3: Estimated probabilities of the intra-individual error for different levels of significance for the variables abscissa and ordinate of the first bifurcation of the STV.
Subject Bifurcation
vein x
Bifurcation
vein y
Bifurcation
artery x
Bifurcation
artery y
1 31 31 31 31
2 32 32 32 32
3 29 29 29 29
4 41 41 41 41
5 30 30 30 30
6 19 19 19 19
7 39 39 39 39
8 28 28 28 28
9 30 30 30 30
10 35 35 35 35
11 8 8 8 8
12 58 58 58 58
13 28 28 28 28
14 17 17 17 17
15 30 30 30 30
16 30 30 30 30
17 12 12 12 12
18 38 38 38 38
19 31 31 31 31
20 31 31 31 31
21 43 43 43 43
22 30 30 30 30
23 34 34 34 34
24 44 44 44 44
25 44 44 44 44
26 41 41 41 41
Total 864 864 864 864
These results indicate a lower probability of error when the data consideredcorrespond to structures such as vessel bifurcations rather than less easily-
defined structures such as the papilla and macula.
1. Boya Vegue J. Atlas de Histología y Organografía microscópica. Médica Panamericana, España, 2002, 393-407.2. Corcóstegui B. El fondo de ojo en la medicina práctica, Espasa, España, 1983, 29-45.3. El Danasoury MA. Prospective bilateral study of night glare after laser in situ keratomileusis with single zone and
transition zono oblation. J Refr Surg 199814 (5): 512-516, 4. Morgan MW. Visual function changes in eyes of aged patients. Vision and Aging: General and Clinical
Perspectives. Ed, A, A, Rosenblom. New York, 1986; 198-200.5. Putz R., Pabst R. Sobotta, Atlas de Anatomía Humana. Tomo I: Cabeza, Cuello, Miembro superior. Médica
Panamericana, Buenos Aires, Madrid, 2006, 362-370.6. Rodríguez S., Smith-Agreda J.M. Anatomía de los órganos del lenguaje, visión y audición. Médica Panamericana,
España, 1999, 250-270.7. Young B., Heath J.W. Wheater´s Histología Funcional. Elsevier España S.A., 2005, 382-391
Material
Figure 2: Focusing the TRC-NW200 retinographer
The retinographer used was theTRC-NW200 by Topcon that does not
require pharmacological mydriasis.Using a corrective lens, images can beobtained in ammetropic eyes of up to -
33 diopters in myopes and +40 inhyperopes.
To further constrain the estimation of this probability of error, we can assume
that the different variables in the individuals behave as independent variables andthen by estimating the probability of an interindividual match for each one(probability of error) the probability of the whole set of variables matching can be
calculated (interindividual probability of error).
The results obtained for this assumption are shown in Table 4.
This study was supported by a formal agreement between the Complutense University
of Madrid and Alta Eficacia Tecnología SL.
MAC X MAC Y BIF X BIF Y PAP X PAP YTOTAL 6 variables
TOTAL 2 variables
Alpha Probability of error
0.01 0.875 0.9 0.0875 0.06667 0.47917 0.63333 0.00139 0.00580.02 0.85 0.85833 0.07917 0.05833 0.43333 0.5875 0.00086 0.00460.05 0.79167 0.7875 0.07083 0.05 0.3375 0.52083 0.00039 0.0035
Tabe 4: Estimated probabilities of the inter-individual error for different levels of significance assuming the independence of variables.
The resolution of the image captured by the CCD camera is 3.15 million pixels.The angle captured is 45º for pupils larger than 4.00 mm and 30º for pupil sizes of
3.70 mm.
Non validretinography
Valid retinography
Pupil diameter <3.70 mm
Pupil diameter >3.70 mm
Figure 5: Artery and vein patterns of the retina in the left and right eyes
Figure 3: Example of the quality of the images obtainedFigure 4: Correct position of the subject for image capture. The
outer corner of the eye is aligned with a line in the instrument.
