In the second problem, automated 3D modeling and quantification of corneal graft detachment after DSAEK based on AS-OCT images, an image processing methodology is developed to detect the region of detachment between the donor lens and the host cornea, for detected the nature of the detachment (communicating or non-communicating) and form the full 3D cornea using AS-OCT images after DSAEK surgery. In the third problem, Automated Detection of Anterior Chamber Depth Using Pentacam Camera Images, an image processing methodology is developed to detect anterior chamber depth not only at the central location but also at different locations.
Introduction
Introduction
The slit-lamp camera, an established ophthalmic device, is primarily used for analyzing the anterior segment of the eye. It is used along with a biomicroscope for observing the eye through it.
Problem statement
Zoom systems are expressed by their focal length range or magnification range. An afocal zoom system [6, 7] is nothing but a zoom system that has an infinite focal length throughout the entire zoom range.
Methods
The magnification of the Galilean type system can be varied by changing either the focal length of the objective lens or the eyepiece. The afocal condition must be maintained throughout the zoom range of the system, i.e. the separation must be equal to the sum of the focal points.
Results
Discussion and future work
It can be seen from the results that the distortion values of the complete biomicroscope are less than 1% in all the configurations, the spot size is also included in the diffraction limit, and the Seidel aberration values are much smaller. Quality assessment, comparison with other zoom systems and real-time design will follow the optimization. Finally, the addition of autofocus lens (readily available) as an objective lens to the fully optimized system will eliminate the patient's chin rest position.
It is collaborative work with Kiran Kumar Vupparaboina, a version of which is present in his thesis.
Introduction
Against this background, we attempted 3D visualization and quantification of donor lenticule detachment (DLS) using only four AS-OCT radial B-scans to aid clinicians in accurate postoperative assessment. The detachment can be communicative (ie, in contact with the aqueduct) or non-communicative (ie, not in contact with the aqueduct). It would be ideal if the specialist also knew the thickness profile and the volume of the deviation, so that he could correctly decide whether further surgery is necessary.
Furthermore, the recent introduction of the intraoperative OCT (iOCT) really helps in performing lamellar keratoplasties more accurately with real-time visualization of the corneal layers [20]. To this end, the ophthalmologist makes a subjective assessment of the type and degree of separation by examining the TD-OCT radial scans. Based on these four 2D sections, the clinician actually mentally synthesizes the 3D topological variation of the detachment.
To this end, it is imperative to develop an automated tool that allows clinicians to visualize and quantify topographic variation of the detachment. In particular, such a tool should accurately detect the type of detachment (either communicative or non-communicative) and estimate the thickness profile and volume of the detachment.
Methods
- Data acquisition
- Proposed Methodology
- Reference Performance: Inter-Observer Repeatability
- Algorithmic Performance Criteria
As a result, the detachment and the background regions seem to acquire similar intensity profiles, making the desired segmentation of the detachment region difficult. Next, corneal outer border is detected by finding the first significant intensity gradient while scanning each column of the binarized image from top to bottom (Fig. 3.3b). Next, in the same vein as earlier, we obtain the initial estimate of the corneal outer boundary by scanning each column of the image (Fig. 3.4c).
However, this estimate sometimes contains spurious peaks/dips, especially at the central location, due to high pixel reflectance (Fig. 3.5a–b). In such cases, we detected the location and extent of the drop and then performed linear interpolation to obtain the desired inner corneal border (Figure 3.5c). With this in mind, we use post-processing to identify significant spots that lie close to the central axis (Figure 3.3k) of the largest connected component within the extent of the donor lenticule (Figure 3.3l).
To this end, we start by delineating the recipient's and the donor's corneal tissues in 2D in addition to the aforementioned detachment region in each of the four radial scans. To this end, we first identify in each of the four radial scans the corneal peak, which divides each delineated area into two halves (Fig. 3.6c).
Results
- Thickness estimates of recipient cornea and donor lentic- uleule
- Volume estimates and overall thickness profile
Consequently, referring to inter-observer repeatability, we obtained QMDC = 1.40 and QCVDC = 1.57 (Table 3.2), which are close to one, indicating the effectiveness of the proposed algorithm. Consequently, we algorithmically measured the thickness of recipient cornea and donor lenticulum at three locations, namely at the central location of the recipient cornea and at peripheral locations 1626.09 µm to the left and right of the central location. In particular, these estimates are calculated along the inward normals at the respective locations of the recipient cornea (see Fig. 3.11).
In this context, we mapped the detachment as well as the corneal regions in 3D, and the volume of the detachment area is obtained. As presented in Table 3.4, the algorithmic volume estimate for our data set ranges between 1.61 mm3 and 15.04 mm3 with a mean of 5.75 mm3 and standard deviation of 3.58 mm3, which closely matches the corresponding reference manual estimates (averaged over four independent estimates made by two different observers) ranging between 1.41 mm3 and 17.04 mm3 with a mean of 6.07 mm3 and standard deviation 4.05 mm3, respectively. In particular, the mean and standard deviation of volume estimates obtained by the proposed method are in close agreement with respective estimates obtained by manual methods, indicating the robustness of the algorithm.
Further, the correlation coefficients between the detachment volume measurements obtained by the manual and proposed methods were observed to be 97.43% which depicts the robustness of the proposed method. The detachment profiles of the six representative subjects considered earlier (see Fig. 3.9) are shown in Figs. 3.13.
Discussion
In addition, it would allow clinicians to measure (and possibly predict) the progression of detachment. At the same time, we plan to study the impact of detachment on visual acuity. Finally, we envision that the proposed algorithm will be used in intraoperative OCT to allow real-time quantitative assessment of graft detachment.
Sobel operator, Roberts operator and Prewitt operator are few examples of the partial differential operators. Here, β and c are thresholds that control the sensitivity of the line filter to the benchmarks, which can be chosen experimentally for specific applications. Accurate and automatic detection of the depth of the anterior chamber has many clinical applications.
This study is used for the automatic detection of the anterior chamber depth of the eye at 7 different positions (one central and 6 peripheral positions) using Rotating Scheimpflug Imaging (Pentacam) system. As the depth of the anterior chamber is the distance from the endothelial layer of the cornea to the anterior capsule of the crystalline lens and iris.
Introduction
Detection and segmentation of anterior chamber parts (cornea, iris and lens) will be the basis of consideration for the depth extraction. By segmenting the image, depth is extracted at different locations, which even extends to 3D anterior chamber imaging. In most studies, the authors wanted to check the applicability of the Rotating Scheimpflug Imaging System (Pentacam) in the calculation of the anterior chamber depth, where they used manual segmentation to check that the weather obtained results are within clinically acceptable levels or not[44, 45].
It has also been found that the anterior chamber depth values obtained by the optical imaging systems such as Orb-scan and Pentacam are within clinically acceptable levels[44]. This study is used for automated detection of anterior chamber depth of the eye in 7 different positions (one central and 6 peripheral positions) using rotating Scheimpflug Imaging (Pentacam) system images. A total of 1225 images of the anterior eye segment of 49 patients are used (25 images angle images of each eye).
Complete automatic methodology that does not depend on any other tools (such as laser beam) for detection of the depth, it will only use the anterior eye segment image. And unlike other methods that will only calculate the depth at a central location, this method allows you to calculate anywhere.
Methods
- Data acquisition
- Proposed Methodology
The anterior segment images (Fig. 4.2.a) taken from Pentacam are generally distorted by speckle noise. The cornea is the largest of all other components in the anterior chamber, so the connected component is used to extract the largest connecting component of the enhanced image as cornea (Fig. 4.3.b). Expand the corneal area (Fig. 4.3.d) and multiply it by the log-linear transformed image, which is nothing but extracting the corneal area from the log-linear transformed image (Fig. 4.3.e).
Then extract the inner border of the resulting image and adjust the circle to form a smooth inner border [32] (Figure 4.3.e). The relative position of the iris and the lens is the basis for perception and segmentation. Thus, the location of the vertical line is extracted using morphological operations, image closure (Figure 4.4.a) and connected components (Figure 4.4.b).
The part touching the vertical line is taken out and is none other than the lens of the eye (Fig. 4.4.e) and its upper edge is labeled (Fig. 4.4.f). Since the detection of the entire lens area of the anterior segment is difficult even manually, we separate the lens and detect only the upper edge.
Results
Since the resolution of the image is 13.34 µm / pixel in horizontal direction and 21.7391 µm / pixel in vertical direction.
Conclusion and Future Work
In addition to 2D, this can be extended to 3D formation of the anterior chamber with eye moment estimation and compensation. Part of the designed system can also be inserted into a discrete zoom system and converted into a continuous zoom system. This work not only helps the tester to precisely select the area of interest, but also eliminates the requirement for a chin rest chamber.
Using the proposed methodology in the second problem, the location of the detachment in the AS-OCT and the nature of detachment (communicative or non-communicative) can be detected. In the future, the tracking of the movement of the eye and the movement compensation will improve the results. But assessment of the accuracy of the algorithm must be done with respect to the manual segmentation and depth calculations.
Eliceiri, “The imagej ecosystem: an open platform for biomedical image analysis,” Molecular Reproduction and Development, vol. Wong, “A new method for gray level thresholding using histogram entropy,” Computer vision, graphics, and image processing, vol.
