Purchase this article with an account.
Charly Maindron, Matthieu Le Loir, Beatrice Cochener, Mathieu Lamard; Automated correction of 3D acquisitions using a high frequency rotative ultrasound probe. Invest. Ophthalmol. Vis. Sci. 2013;54(15):5517.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Phakic implantation in the posterior chamber appears to be the best operating procedure for high ametropia correction. Its optimal clinical use is however limited by the lack of knowledge regarding the positioning in the sulcus, which exact 3D shapes and dimensions are unknown. These could potentially be obtained by the automatic segmentation of reconstructed ultrasound (US) images. In order to optimize this step, we have developed an algorithm allowing for the improvement of reconstructed data.
24 myopic eyes (mean MRSE: -8.54D) of 13 patients and 16 hypermetropic eyes (mean MRSE: +8.12D) of 8 patients underwent a posterior phakic implantation by one surgeon using ICL V4 (STAAR Surgical®) with a “white to white”-based sizing method. Data acquisition was carried out using a 50Mhz linear scanning US probe (Aviso©, Quantel Medical®) associated with a motor, able to capture 30 coaxial slices along a 180° rotation axis in 8 sec. This acquisition duration is associated with unavoidable movements of the physician and/or the patient, leading to decreased quality of the reconstructed data. The 30 images were therefore processed prior to automatic segmentation, in order to optimize the subsequent automatic registration using a novel fitting algorithm. As the probe rotates around a fixed axis, the structures within each image were assumed to be the same. One pixel wide columns at the center of each image, corresponding to this rotation axis, were thresholded, compared and then registered to optimize their positioning.
Processing of an entire 30 slices volume was instantaneous so it could be used in clinical practice. A mean 0.12mm (11 pixels) (±0.28mm) vertical fitting was observed over 92 acquisitions. Registration results were visually assessed by experts to be accurate and better adjusted to the fitted data. The automatic segmentation performed on the reconstructed data using the proposed fitting method reduced Haussdorff distance between manual and automatic posterior chamber reconstruction by 8%.
The proposed fitting algorithm allowed a significant improvement of the reconstructed data, leading to a more accurate and faster subsequent automatic segmentation step. Future work will consist in introducing within our model the laws describing mechanical properties of the eye structures in order to simulate the eye structural modifications introduced by the phakic IOL positioning.
This PDF is available to Subscribers Only