Investgatve Ophthalmology & Vsual Scence, Vol. 29, No. 8, August 1988 Copyrght Assocaton for Research n Vson and Ophthalmology Vdeographc Measurements of Optc Nerve Topography n Glaucoma Joseph Caprol and Joseph M. Mller Topographc measurements of the optc nerve head were made wth computerzed vdeographc mage analyss (Rodenstock Analyzer) n one eye each of 36 normal controls, 41 glaucoma suspects and 46 glaucoma patents matched for age. Glaucoma suspects had elevated ntraocular pressures and normal vsual felds n both eyes. Glaucoma patents had typcal vsual feld defects. Dsc measurements were corrected for the optcal dmensons of ndvdual eyes. One-way analyss of varance revealed statstcally sgnfcant dfferences among the dagnostc groups for cup-dsc rato (P = 0.0006), dsc rm area (P < 0.0001) and cup volume (P = 0.0001). Mean (±SEM) dsc rm area was 1.14 ± 0.04 mm 2 for controls, 1.10 ± 0.04 mm 2 for glaucoma suspects and 0.87 ± 0.05 mm 2 for glaucoma patents. Mean (±SEM) optc nerve cup volume was 0.35 ± 0.02 mm 3 for controls, 0.44 ± 0.04 mm 3 for glaucoma suspects and 0.60 ± 0.05 mm 3 for glaucoma patents. Planmetrc measurements of dsc rm area were made from manual tracngs of stereoscopc dsc photographs of the same eyes. There was a statstcally sgnfcant correlaton between the computerzed vdeographc measurements and the manual photographc measurements of dsc rm area (r = 0.73, P < 0.0001). The broad range of values for these optc nerve structural parameters n normal eyes and ther overlap wth values n glaucomatous eyes prevents ther use to relably predct whch patents are normal and whch have glaucomatous vsual feld loss. New parameters are requred to fully descrbe the depth nformaton generated wth new quanttatve technques. Invest Ophthalmol Vs Sc 29:1294-1298,1988 Some clncal studes suggest that structural changes of the optc nerve head precede measurable vsual feld abnormaltes n early glaucoma. 1 " 7 Consderable loss of optc nerve fbers can occur n patents wth elevated ntraocular pressure before vsual feld abnormaltes are detected. 8 The earlest dsc changes n patents wth glaucoma probably cannot be adequately dentfed by clncal estmates of the cup-dsc rato. 9 Careful examnaton of stereoscopc dsc photographs seems to be the best routne way to detect subtle nterval changes of the optc dsc. 3 ' 4 ' 7 The qualtatve and subjectve nature of these comparsons has ntated a search for more quanttatve methods. From the Glaucoma Servce, Yale Unversty School of Medcne, Department of Ophthalmology and Vsual Scences, New Haven, Connectcut. Supported n part by grants from The Texaco Phlanthropc Foundaton, Research to Prevent Blndness Inc., The Robert Leet and Clara Guthre Patterson Trust, and the Connectcut Lons Eye Research Foundaton Inc. Dr. Mller was supported by NIH Tranng Grant 5 T32 EY-07000-10. Submtted for publcaton: July 29, 1987; accepted March 28, 1988. Reprnt requests: Joseph Caprol, MD, Yale Unversty School of Medcne, Department of Ophthalmology and Vsual Scences, 330 Cedar Street, New Haven, CT 06510. We used computerzed mage analyss (Rodenstock Analyzer) to measure cup-dsc rato, dsc rm area, and dsc cup volume n age-matched normal controls, glaucoma suspects and n patents wth vsual feld loss from glaucoma. The automated measurements of dsc rm area were also compared wth manual measurements of dsc rm area made from stereoscopc dsc photographs. Materals and Methods Patents of the Yale Glaucoma Servce were used n ths study. All patents over the age of 40 who were evaluated for glaucoma before December 1, 1986 and who had automated permetry (Octopus Program 32; Interzeag, Schleren, Swtzerland), stereoscopc optc dsc photographs, and computerzed mage analyss (Rodenstock Analyzer; Rodenstock Instrumente, Gmbtl., Munch, West Germany) were ncluded. Computerzed mage analyss was performed n patents wth reasonably clear meda (20/40 or better) who could be dlated to 5 mm or more. Those who had elevated ntraocular pressure (consstently >21 mm Hg n both eyes) and unequvocally normal vsual felds n both eyes (no loss compared to agematched controls) were consdered glaucoma suspects (GS). Those wth elevated ntraocular pressures (or a hstory of elevated ntraocular pressure before 1294
No. 8 OPTIC NERVE TOPOGRAPHY IN GLAUMA / Coprol and Mller 1295 treatment) and typcal glaucomatous vsual feld defects were consdered glaucomatous (GL). Typcal glaucomatous defects were defned as at least three contguous ponts wth ten decbel loss or greater compared to age-matched controls n the superor or nferor Bjerrum areas, or a 10 decbel dfference across the horzontal nasal mdlne at two or more locatons. The vsual feld mean defect and corrected loss varance was calculated for each feld. 10 Normal control subjects (NL) had no hstory of eye dsease and were recruted from hosptal staff and spouses or frends of patents; none had presented to the Eye Center for evaluaton. Those ncluded were over the age of 40 and had a normal eye examnaton and normal vsual felds tested wth automated threshold permetry. Informed consent was obtaned from each person after photographc procedures were fully explaned. A computerzed mage analyss system (Rodenstock Analyzer) was used to make quanttatve depth measurements of the optc nerve head. Descrptons of ths nstrument and the reproducblty of ts measurements have been publshed."" 13 The optcal head provdes a smultaneous stereoscopc vdeographc mage of the optc dsc wth an mage-ntensfed vdeo camera. A frame grabber (256 X 256 X 8 bt buffer) dgtzes and freezes fundus mages for evaluaton by the operator. Images may be rejected or saved on a floppy dskette or fxed dsc. To obtan depth measurements across the optc nerve surface, a set of seven parallel strpes s projected onto the fundus; two such mages are processed to make depth measurements and provde 14 vertcal strpes along whch measurements are made. Approxmately nne to ten of these span an average-szed optc nerve head. The strpes provde features on otherwse featureless areas of the optc dsc so that effectve mage analyss can be performed. The nstrument smultaneously records a par of stereoscopc dsc mages wth strped patterns whose deformatons contan depth nformaton. The computer selects correspondng lne segments wthn the two stereoscopc half-mages and computes ther cross-correlaton functon to determne depth. These depth measurements are processed to extract structural parameters after the dsc edge s defned. The dsc edge s defned by an ellpse whch s postoned by the operator. A computer algorthm dentfes the margn of the cup as a locus of 360 ponts, one on each of 360 radal profles whch le 150 nm below and nearest the dsc edge. The dsc rm area s the area between the cup rm margn and the dsc edge. The cup-dsc rato s the average of. all cup-dsc ratos calculated for each of the 360 radal profles. The optc nerve cup volume s the volume below an magnary plane constructed at the average level of the perpapllary retna and bounded by the dsc edge. All dmensons are corrected for optcal magnfcaton or mnfcaton wth sonographc measurements of the axal eye length or wth keratometrc and refractve measurements. 14 Stereoscopc optc dsc photography was performed wth a Zess fundus camera equpped wth a X2 adaptor. Photographs were taken on Kodachrome slde flm. The best stereoscopc par was selected for each dsc, and the rght-hand stereophotograph was duplcated and projected onto a sheet of paper. The dsc edge and perceved cup edge was traced by one of us (JC) whle the stereoscopc par was vewed; only depth clues were used, and color dfferences were gnored. The locaton of the cup edge was defned as beng at the level of the nternal scleral openng when vewed stereoscopcally. The dsc area was set to equal the prevously calculated dsc area for each eye as descrbed above. A computerzed dgtzng pad was used to measure the dsc rm area whch was corrected to an absolute value n mm 2. Statstcal analyss was performed wth the goodness offttest, student t-test, one-way analyss of varance (ANOVA), and Duncan's multple range test. 15 Results Statstcs for age, refractve error, optc dsc parameters and vsual feld ndces are gven n Table 1. Analyss of varance demonstrated no statstcally sgnfcant dfferences for age or refractve error (sphercal equvalent) among the groups. The mean total dsc area dd not dffer among the groups (ANOVA, P = 0.35). ANOVA revealed statstcally sgnfcant dfferences among the groups for cup-dsc rato (P = 0.0006), dsc rm area (P < 0.0001), and cup volume (P = 0.0001). The Duncan multple range test was used to test for smultaneous statstcally sgnfcant dfferences (P < 0.05) between the ndependent means of each dsc parameter among the patent groups. Mean cup-dsc rato, mean cup volume and mean dsc rm area were sgnfcantly dfferent between the NL and GL group, and also between the GS and GL group. There s consderable overlap of the measurements for each of these structural parameters among the patent groups (see Fgs. 1-3). There was a postve correlaton between dsc rm area measurements performed manually and those performed wth computerzed mage analyss (Fg. 4). Lnear regresson analyss provded a best ft wth slope = 0.87, y-ntercept = 0.27 mm 2, and correlaton coeffcent (r) = 0.73 (P < 0.0001). The correlaton coeffcents for each of the patent groups taken separately were: 0.56 for NL, 0.66 for GS and 0.72 for GL. Eght outlers were dentfed from the scatterplot
1296 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / August 1988 Vol. 29 Table 1. Statstcs for age, refractve error, optc dsc parameters and vsual feld ndces* Normal (n = 36) Glaucoma suspect (n = 41) Glaucoma (n = 46) Age (years) Refractve error (sphercal equvalent, dopters) Optc Dsc Dsc area (mm 2 ) Cup-dsc rato Dsc rm area (mm 2 ) Dsc cup volume (mm 3 ) Vsual feld Mean defect (db) Corrected loss varance (db 2 ) 62.8 ± 1.6 0.2 ±.3 1.71 ±0.05 0.49 ± 0.03 1.14 ±0.04 0.35 ± 0.02 59.7 ± 1.5-0.3 ±.4 1.79 ±0.07 0.54 ± 0.03 1.10 ±0.04 0.44 ± 0.04-0.4 ±0.3 4.9 ± 1.9 63.7 ± 1.3-0.3 ±0.4 1.70 ±0.05 0.64 ± 0.03 0.87 ± 0.05 0.60 ± 0.04 7.4 ± 1.0 22.8 ±3.9 1 Data are presented as the mean ± SEM. (Fg. 4); the rato of the manual to the computerzed measurements was less than 0.6 for these eyes. Seven eyes were from subjects wth glaucoma, and one was from a glaucoma suspect. All eght had the sngle characterstc of a concave or flat dsc rm whch sloped contnuously upward to the dsc edge. The sloped dsc rm n these cases was not ncluded n the manual rm area measurement, but portons wthn 150 xm n depth from the dsc edge were ncluded n the computerzed measurement. For ths reason, the dsc rm area determned by mage analyss was larger than that measured wth the manual technque. Dscusson Ophthalmologsts have tradtonally used estmates of cup-dsc rato to evaluate the optc nerve for glaucomatous damage. Bengtsson showed that the dameter of the optc nerve cup depends on the sze of the optc dsc, and suggested that the breadth of the dsc rm mght be a more useful measure of glaucomatous optc nerve damage. 16 Radus and Pederson 17 found a correlaton between the amount of rm tssue and the number of axons present n the optc nerves of prmate eyes wth expermentally nduced glaucoma. Absolute measurements of optc nerve topography may provde an estmate of the number of nerve fbers contaned n the optc nerve head and may assst the detecton of early glaucomatous nerve damage. The dsc rm s narrowed n patents wth vsual feld defects from glaucoma. 18 Balazs et al 19 made absolute measurements of the dsc rm, whch the authors termed "neuroretnal rm." The locaton of the dsc rm and dsc edge were traced on a projected mage whle a correspondng stereoscopc par was vewed. The traced dsc rm area was measured wth a computerzed planmeter. Enlarged black and whte 1.0 2.0 o O 0.5 Q I Q. O CM e E HI < 1.0 o V* 4- Fg. 1. The dstrbuton of cup-dsc ratos for normal subjects (n = 36), glaucoma suspects (n = 41) and patents wth glaucoma (n = 46). There are statstcally sgnfcant dfferences between the Fg. 2. The dstrbuton of dsc rm areas for normal subjects (n = 36), glaucoma suspects (n = 41) and patents wth glaucoma (n = 46). There are statstcally sgnfcant dfferences between the
No. 8 OPTIC NERVE TOPOGRAPHY IN GLAUMA / Coprol and Mller 1297 stereoscopc prnts were used by Araksnen and coworkers to make smlar measurements n normal, glaucoma suspect and glaucoma patents. 20 The perceved poston of the cup edge was traced and corrected area measurements were made. Means (±SEM) for dsc rm area were measured n ths way were 1.40 ± 0.03 mm 2 n 33 normals, 1.31 ± 0.04 mm 2 n 25 glaucoma suspects, and 0.89 ± 0.05 mm 2 for 38 glaucoma patents. There were statstcally sgnfcant dfferences between the means, but the values overlapped consderably among the patent groups. Measurements of cup volume are not possble wth ths technque. We used a computerzed system for mage analyss of smultaneous stereoscopc vdeographc mages of the optc nerve head (Rodenstock Analyzer) to make optcally corrected measurements of dsc sze, cupdsc rato, dsc rm area and dsc cup volume. The reproducblty of the measurements (average coeffcent of varaton) obtaned wth the verson of the system used n ths study s 6.1% for cup dsc rato, 6.3% for dsc rm area, and 9.2% for cup volume. 12 There were sgnfcant dfferences between mean measurements of each of these parameters between normals and glaucoma patents, and between glaucoma suspect and glaucoma patents, but not between normals and glaucoma suspects. The broad range of values for dsc rm area n normal eyes and ther overlap wth values n glaucomatous eyes prevents the use of ths parameter to relably predct whch patents are normal and whch have glaucomatous feld loss, and seems only slghtly better than measurements of cup-dsc rato or cup volume n ths regard. The cup margn as defned by the computer program does not necessarly correspond to the clncal percepton of the cup margn. There are some dscs for whch the measurement of dsc rm area wth the automatc and manual technques dffers. Ths s not surprsng, snce the defntons of the cup margn wth the two methods are qute dfferent. Subjectve percepton of the cup margn s dffcult to standardze, and may nvolve such thngs as the rate of change of slope, depth, locaton and pattern of vessels, vsblty of the lamna, and sometmes dfferences n color. Automatc localzaton of the cup edge can be standardzed, and n ths analyss was defned as lyng 150 nm below the dsc margn. Eyes n whch there was a dscrepancy between the manual and automatc measurements had characterstcally concave or flat dsc margns whch sloped upward to the dsc edge. In these cases, areas of the cup edge were manually marked concdent wth the dsc edge, whle the automatc technque ncluded some portons (those wthn 150 ^m n depth from the dsc edge) of the dsc EE LU 1.0 0.5 0 - t tb r * Fg. 3. The dstrbuton of cup volumes for normal subjects (n = 36), glaucoma suspects (n = 41) and patents wth glaucoma (n = 46). There are statstcally sgnfcant dfferences between the rm substance n the calculaton for dsc rm area. Nevertheless, the overall correlaton between the manual and automatc technques for measurements of dsc rm area was qute good (Fg. 4). The structural parameters reported here have been derved from quanttatve measurements of surface contour and were orgnally calculated to conform to LU 2.0 1 0.5" DISC RIM AREA (mm 2 ) * r t : NL : GS * : GL 0 0.5 1.0 1.5 2.0 MPUTERIZED IMAGE ANALYSIS Fg. 4. Measurements of dsc rm area obtaned wth manual tracng and planmetry are plotted aganst those obtaned wth computerzed mage analyss (Rodenstock Analyzer). There s a statstcally sgnfcant correlaton between these two sets of measurements (r = 0.73, P < 0.0001). The dotted lnes ndcate the 95% confdence nterval for the lnear regresson. There are eght outlers for whch the dsc rm area measurement wth the manual technque was much smaller than wth the computerzed technque. Please see text for detals.
1298 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / Augusr 1988 Vol. 29 our present clncal descrptons of the optc nerve head. However, these parameters now seem nadequate to detect small changes n surface contour of the dsc and surroundng nerve fber layer. It may be necessary to abandon these summary structural parameters so that measurements of surface contour of the dsc and nerve fber layer may be used to best advantage. Comparsons of sequental quanttatve measurements of surface contour may more effectvely detect early structural damage from glaucoma. Key words: optc nerve, glaucoma, mage analyss, cup-dsc rato, dsc rm area, dsc cup volume Acknowledgments The authors thank Maureen Roche and Pamela Ossoro for techncal assstance. References 1. Armaly MF: The correlaton between appearance of the optc cup and vsual functon. Trans Am Acad Ophthalmol Otolaryngol 73:898, 1969. 2. Aulhorn E and Harms H: Papllen veranderung und geschtsfeld storung bem glaukom. Ophthalmologca 139:279, 1960. 3. Odberg T and Rse D: Early dagnoss of glaucoma. Acta Ophthalmol 63:257, 1985. 4. Pederson JE and Anderson DR: The mode of progressve dsc cuppng n ocular hypertenson and glaucoma. Arch Ophthalmol 98:490, 1980. 5. Pckard R: A method of recordng dsc alteratons and a study of the growth of normal and abnormal dsc cups. Br J Ophthalmol 7:81, 1923. 6. Read RM and Spaeth GL: The practcal clncal apprasal of the optc dsc n glaucoma: The natural hstory of cup progresson and some specfc dsc-feld correlatons. Trans Am Acad Ophthalmol Otolaryngol 78:255, 1974. 7. Sommer A, Pollack I, and Maumenee AE: Optc dsc parameters and onset of glaucomatous feld loss. Arch Ophthalmol 97:1444, 1979. 8. Qugley HA, Addcks EM, and Green WR: Optc nerve damage n human glaucoma: III. Quanttatve correlaton of nerve fber loss and vsual feld defect n glaucoma, schemc optc neuropathy, paplledema, and toxc neuropathy. Arch Ophthalmol 100:135, 1982. 9. Lchter PR: Varablty of expert observers n evaluatng the optc dsc. Trans Am Ophthalmol Soc 74:532, 1976. 10. Flammer J, Drance SM, Augustny L, and Funkhauser A: Quantfcaton of glaucomatous vsual feld defects wth automated permetry. Invest Ophthalmol Vs Sc 26:176, 1985. 11. Cornsweet TN, Hersh S, Humphres JC, Beesmer RJ, and Cornsweet DW: Quantfcaton of the shape and color of the optc nerve head. In Advances n Dagnostc Vsual Optcs, Brenn GM and Segel IM, edtors. New York, Sprnger-VerlagNYInc, 1983, pp. 141. 12. Caprol J, Klngbel U, Sears M, and Pope B: Reproducblty of optc dsc measurements wth computerzed analyss of stereoscopc vdeo mages. Arch Ophthalmol 104:1035, 1986. 13. Mkelberg FS, Douglas GR, Schulzer M, Cornsweet NN, and Wjsman K: Relablty of optc dsk topographc measurements recorded wth a vdeo-ophthalmograph. Am J Ophthalmol 98:98, 1984. 14. Lttmann H: Zur Bestmmung der wahren Grosse enes Objektes auf dem Hntergrund des lebenden Auges. KJn Monatsbl Augenhelkd 180:286, 1982. 15. Mller RG: Smultaneous Statstcal Inference. Second Edton. New York, Sprnger-Verlag, 1981, pp. 81-90. 16. Bengtsson B: The varaton and covaraton of cup and dsc dameters. Acta Ophthalmol 54:804, 1976. 17. Radus RL and Pederson JE: Laser-nduced prmate glaucoma: II. Hstopathology. Arch Ophthalmol 102:1693, 1984. 18. Betz PH, Camps F, Collgnon-Brach J, Lavergne G, and Weekers R: Bometrc study of the dsc cup n open-angle glaucoma. Graefes Arch Cln Exp Ophthalmol 218:70, 1982. 19. Balazs GA, Drance SM, Schulzer M, and Douglas GR: Neuroretnal rm area n suspected glaucoma and early chronc open-angle glaucoma. Arch Ophthalmol 102:1011, 1984. 20. Araksnen PJ, Drance SM, and Schulzer M: Neuroretnal rm area n early glaucoma. Am J Ophthalmol 99:1, 1985.