New Method to Describe Cone Shape in Keratoconic Corneas

Lopes, Bernardo T ORCID: 0000-0002-8489-3621, Abass, Ahmed ORCID: 0000-0002-8622-4632, Eliasy, Ashkan ORCID: 0000-0002-4473-1900, Zhang, Haixia and Elsheikh, Ahmed ORCID: 0000-0001-7456-1749
(2021) New Method to Describe Cone Shape in Keratoconic Corneas. In: ARVO 2021 Annual Meeting, 2021-5-1 - 2021-5-7, Rockville, Md. USA.

[img] Text
Poster_ARVO_2021_pdf.pdf - Author Accepted Manuscript

Download (598kB) | Preview


Purpose: To introduce and validate a novel method to describe the cone features in keratoconic corneas. Methods: Corneal anterior and posterior surfaces were described using a spherical coordinate system to generate a new spherical height map to allow the detection of the cone apex. Cone boundaries were objectively estimated using second derivatives of spherical height in an iterative process. Corneal topography exams of 309 keratoconic patients with different disease severities were evaluated with the new automated method as its first validation. In addition, 12 cornea specialists blindly evaluated the tangential and elevation maps relative to the best-fit sphere of 6 patients. Their estimations were cross-checked and compared with the results of the new automated method in order to evaluate the subjective variability and provide a second validation of the method. Results: The main cone features in the anterior and posterior surfaces were evaluated in the large clinical dataset. There was a strong correlation between the cone height and the disease severity in both surfaces (R=0.71, p<0.01), while the disease stage did not show a significant correlation with cone area in any of the surfaces (R=0.01, p=0.77). The height of the posterior cone was, on average, larger than the corresponding anterior cone height by 37 ± 24 μm (0–158). In relation to the experts’ assessment, there was high inter-subject variability, up to 55%, among experts’ estimations of the cone area and low intra-subject agreement in cone apex location using different maps (p<0.05). However, there was no statistically significant difference between the automated estimation and the specialists’ estimations in both maps (p>0.05). The cone boundaries estimated by the automated method were within the range of the specialists’ estimations in all cases. Conclusions: An objective automated method able to determine the cone’s 3D features was developed and validated in a large clinical dataset and against worldwide corneal specialists’ estimations. The method’s results were in agreement with disease severity and independent of the subject variability observed among different experts. The objective method provided a reliable and unique evaluation of keratoconus features that is independent of maps’ type or color scale.

Item Type: Conference or Workshop Item (Unspecified)
Divisions: Faculty of Science and Engineering > School of Engineering
Depositing User: Symplectic Admin
Date Deposited: 19 Apr 2021 09:42
Last Modified: 18 Jan 2023 22:52
Related URLs: