Abstract
Purpose: :
We propose a new method to perform the semi-automatic 3D segmentation and reconstruction of the anterior segment (AS) using high frequency ultrasound (HFUS) scans. The main application aimed at by this development is the phakic IOL sizing based on the 3D morphological quantification of the AS : ATA, STS …
Methods: :
High frequency ultrasound (HFUS) data were acquired by a linear 50 MHz probe ( Aviso from QUANTEL MEDICAL), using freehand scanning. An active localization system (made of an infrared camera associated with sensors placed on the probe) was used to spatially register each scan in a common 3D coordinates system. The steps of the AS segmentation were defined as follows: 1 - 3D reference models (one for each organ: the iris, sulcus, lens, ciliary body and the cornea) were previously created using manual US image datasets segmentation. 2 - The 3D models were globally registered to the acquired data using rigid transformations based on landmarks. 3 - A local refinement was carried out using a locally adaptive elastic registration of the models to the data. 4 - A final manual control / correction ensures the accuracy of the overall procedure. Tests were performed on 9 volumes (containing from 30 to 45 slices) acquired in the same conditions, during the preoperative check up.
Results: :
Our method is much less time consuming than manual segmentation: it took 192 min for the trained expert to segment a medium-sized dataset (30 scans), whereas our entire proposed procedure for 3D anterior segment reconstruction required about 21 ± 7 min only. The morphological values of the ATA and STS 3D reconstruction were extracted for all the meridians. The ATA had an average minimum and maximum radius of 5.38 ± 0.15 and 6.74 ± 0.21 mm respectively, and an average area of 108.94 ± 4.6 mm2. For the STS, an average minimum and maximum radius of 4.93 ± 0.22 and 6.71 ± 0.38 mm was determined, and its average area was 102.61 ± 5 mm2.
Conclusions: :
The proposed method offered an accurate and fast 3D segmentation of the ocular anterior segment organs. Future work will aim at reducing the required time and human intervention for each patient datasets processing for better automation. In addition new morphological values adapted to the phakic IOL sizing will be investigated.
Keywords: anterior segment • image processing • refractive surgery: phakic IOL