An  automated system of measuring decentration of anterior segment structures from  geometric central  axis 

Tariq M Aslam*,  Abha Gupta*, Chris Rose**, .

*          Manchester Eye Hospital

**        Manchester University.

  Purpose 

To develop and test a system of automated analysis of amount of decentration of anterior segment structures from a geometric axis defined either by pupil or limbal landmarks. The system should be openly available and accessible for use without specialised equipment.

 Background 

Accurate and precise measurement of an object’s decentration in ophthalmic examination is useful for both scientific studies and clinical assessments.[1] In scientific studies, for example the level of decentration may be one of the measured outcome measures of a comparative study. In clinical examination, the level of decentration measured sequentially in time may provide important prognostic information for patients and to plan management.

These measures are not easily done with standard imaging software. For example, intraocular lens or natural lens decentration may require assessment when only a small portion of the perimeter curve is visible. Other objects may require measurement such as pupil decentration from limbus, PCO or  corneal scarring.

Complex systems for measurement of intraocular lens decentration do already exist that are dependent on the analysis of reflected purkinje images.[2, 3] These systems are very accurate and able to give information on tilt as well as decentration. However, they require specialised equipment to measure and analyse data. Also, these systems depend on the formation of reflection purkinje images which may not be adequate when dealing with dense cataracts, corneal haze/ abscesses or other non-reflective bodies.

The system presented allows for measurement of any circular bodies that are visible on digital images. The decentration of the centre of these bodies from a user determined centre of axis ( e.g centre of limbus / pupil ) can be measured. The distance and angle from centre of the body to centre of the axis is calculated. To counter the problems caused by variable magnification, data on distance is presented not only in pixels, but also in relation to a primary constant circular diameter distance that the user marks out. This constant can, for example, be the limbus or diameter of an intraocular lens. If the actual size of this distance can be measured or is known, for example in millimeters, distance in millimetres of decentration can be calculated from the computers output.

 Method  

To comply with accessibility requirements the system was designed to function through analysis of patients standard anterior segment digital images in .jpg form , viewed on a computer running Windows XP. No further specialise components or set up is required. A program was developed using the Matlab program[4] along with the Image processing toolbox ©The Mathworks. The program code was compiled to be able to be used on machines without the Matlab software.

The analysis is based on the user uploading any digital image of the eye to be measured into the program.(Fig. 1)  The user must define three separate circles of interest in this image. Each circle is defined by the user clicking on three points that are visible and deemed to be on the circle perimeter, with each point as far apart as possible from the others. The software calculates a best fit perimeter that connects these three points and also marks out the centre of that circle.

 Firstly the user delineates a reference circle ( Fig. 1 &2 ). The diameter of this circle is used as a constant, and final calculated distances are not only expressed as pixels but also fractions of this constant. As such, a circle with known, fixed, or measurable diameter should be chosen for this first circle, perhaps the limbus of the eye.

The second circle that is delineated is used to calculate the central point from which all distances and angles are measured.( Fig. 3)  This may commonly be of either the limbus or pupil borders. The geometric central point is thus calculated based on limbal or pupil structure.

Finally the perimeter of the object to be measured is delineated.( Fig 4)  The user clicks on three points, diametrically opposed if possible, on the object whose decentration is to be assessed. This object may be the pupil itself ( if being measured relative to limbus) , crystalline lens, intraocular lens, corneal opacity, posterior capsule opacification or other anterior chamber body. The computer again calculates the best fit circle to this final set of points, as well as its centre.

Finally, the level of decentration in pixels and angle in degrees of decentration of this object to the second circle’s centre point is displayed. The distance of decentration is also calculated as a fraction of the reference circle diameter. With appropriated correction this distance can be converted to a millimetre scale if reference circle diameter is known or measurable. If not, it still allows for more precise comparisons between different images of the same eye taken with perhaps slightly different magnifications.

The system was then tested for both reliability and validity.

To test for reliability a user measured images of patients with decentred lenses and IOLs on two separate occasions , with significant time period between measurements. Further reliability studies are planned with more measures and two separate investigators.

To test validity[5], a physical model  artificial eye was used. The centre of the lens and of the pupil were determined geometrically in the model artifical eye. The lens was decentred by set amounts, measured as best as possible with a ruler. Photographic images of the eye were then taken. ( Fig. 5 ) The computer system was used to analyse these images.The calculated decentration in pixels was converted to a millimetre scale using the reference of the corneal limbus whose length was known after also being measured in millimetres. The final distance calculated by the computer was compared with the known values measured by ruler .

  Results  

The compiled system is fully functional on Windows XP, when installed with supplied software. The program is user friendly and  menu driven. The  image is first loaded with a pull down menu. Circle of interest was found to be accurately placed for both object and reference measurement structures. The system was found to be easily adaptable to assess decentration of pupils ,  crystalline lens , intraocular lens and posterior capsule opacification.

The intraindividual reliability has as yet only been measured. However,  this reliability of the system was found to be high. Results are plotted in Fig. 6.

Ample evidence for validity of the program was demonstrated by strong agreement of measurements produced with the program compared to exact known measurements on test model decentred images.( Fig. 7) Variations that were found reflect the inaccurate nature of measurements made on the model eye, especially after fixing components in irregular positions. Further studies will involve more and more accurate measures on the model eye and also further validity studies using specifically created flat paper templates whose centre points and diameter and decentration can be measured with a much higher degree of accuracy.

Conclusions.

This paper presents as system of analysis of decentration of anterior segment bodies. It requires no specialised equipment bar standard digital photography imaging equipment and a computer running Windows XP. The program and necessary files can be installed using the supplied software. The system is shown to be extremely versatile, valid and reliable. Further studies on reliability and validity are required however and these have been planned to extend evidence for the usefulness of this computerised automated measure.

The system can be freely obtained by contacting the author.