Measuring the Ocular Morphological Parameters of Guinea Pig Eye with Edge Detection and Curve Fitting

Abstract

Aim. To identify the guinea pig eyeball with edge detection and curve fitting and devise a noncontact technology of measuring ocular morphological parameters of small experimental animal. Methods. Thirty-nine eyeballs of guinea pig eyeballs were photographed to obtain the anterior and posterior surface; transverse and sagittal planes after the eyeballs were eviscerated. Next, the eyeball photos were input into digital image analysis software; the corresponding photo pixels-actual length ratio was acquired by a proportional scale. The contour lines of the eyeballs were identified by edge detection technology; conic curve fitting was applied to fit the contour line of the eyeball. The maximum and minimum diameters, the horizontal and vertical diameters, eccentricity, tilt angle, cross-sectional area, equatorial circumference, retrobulbar equatorial maximum length, corneal radius of curvature (CRC) in central region, and the whole cornea were calculated according to the geometric principles. The corneal data of in vitro study were compared with the in vivo results. Results. The contour line of the selected guinea pig eye was identified correctly by edge detection. There were no significant differences between anterior and posterior surfaces of one eyeball in the maximum diameters, eccentricity, cross-sectional area, equatorial circumference, and tilt angle ( P > 0.01 ). There were significant differences of eccentricity and CRC between central region and the whole cornea ( P < 0.01 ). There were no significant differences between keratometer in vivo and cornea in vitro. Conclusion. It was feasible to measure an experimental animal eye in a noncontact way. Edge detection and curve fitting technology could accurately evaluate the ocular morphological parameters.

Digital measuring the ocular morphological parameters of guinea pig eye in vivo with Python

“…The actual distance conversion coefficient (Figure 4D) could be obtained by taking points on the graduated scale at 10 mm distance repeatedly and calculating the average value. This above method was the same as the method of actual length-photo pixels conversion in Matlab image in our previous research [12] . Acquisition of ocular edge data manually Ten data points were manually taken in the target contour (corneal edge) to obtain the coordinates.…”

“…Assuming the center of the cornea is a circle, circle fitting was applied, the real curvature radius of the cornea by means of the conversion coefficient could be obtained, just like we reported before [12] (Figure 5A, 5C, 5E).…”

“…To confirm the reliability, the edge detection which reported before was also applied for comparison [12,15] : First, the guinea pig was sacrificed by excessive anesthesia. A point was marked by a glowing needle tip at the top of the corneoscleral limbus before eviscerating the eyeballs.…”

“…However, it is still unsatisfactory not only additional expenditures, but also limited measuring parameters. In previous research in guinea pig, we identified the contours with edge detection, and curve fitting the eyeball with Matlab software, the corneal curvature radius was obtained by means of numerical simulation [12][13][14] (Figure 1). The corneal eccentricity, the long and short axis of the eye and the eye axis were achieved in the same way (Figure 2).…”

Tongue image quality assessment based on a deep convolutional neural network

“…In general, CNN architectures can avoid feature selection manually and automatically extract features, which are key elements to enable the intelligent tongue diagnosis system into TCM clinical practice. Although several previous studies have reported encouraging results using CNN methods to extract tongue image features for tongue colour (tongue body and tongue coating) classification [ 17 – 19 ], tongue image characteristic recognition (tooth-marked tongue [ 20 – 22 ], tongue cracking [ 23 ]), tongue image segmentation [ 24 – 31 ], and clinical application in herbal medicine [ 32 , 33 ], they usually ignore the quality of tongue images or implicitly assume the good quality of tongue images. Thus, the medical application of deep learning methods to the field of tongue diagnosis has not achieved much so far.…”