Acquisition of the curves of the human crystalline lens from slit lamp images: an application of the Hough transform.

To accurately model lens-based functions such as accommodation and image formation on the retina, it is essential to know anterior chamber depth, anterior segment length, lens thickness, and, most importantly, lens curvature both on the surfaces and internally. With the exception of lens curvatures, all these data can be obtained with a high degree of precision by one or more techniques (i.e., A-scan ultrasonography and pachymetry). Lens curvatures can be collected by Scheimpflug slit lamp photography, but the curvature data must be extracted from these images, a problem complicated by both linear and nonlinear image distortion. Previous approaches have involved significant magnification of the image combined with major subjective input and judgment. We present here a computer-based application of the Hough technique for measurement of curvature of lens surfaces observed in Scheimpflug slit lamp photography and related evaluation of (and solutions for) the associated image distortion. Minimal user input is required for successful application of this method, but the time required to obtain a fit for each surface is >1 min. Results obtained by this technique on test images compare favorably with those obtained by independent methods.