Automatic Three-dimensional Detection of Photoreceptor Ellipsoid Zone Disruption Caused by Trauma in the OCT

Abstract

Detection and assessment of the integrity of the photoreceptor ellipsoid zone (EZ) are important because it is critical for visual acuity in retina trauma and other diseases. We have proposed and validated a framework that can automatically analyse the 3D integrity of the EZ in optical coherence tomography (OCT) images. The images are first filtered and automatically segmented into 10 layers, of which EZ is located in the 7(th) layer. For each voxel of the EZ, 57 features are extracted and a principle component analysis is performed to optimize the features. An Adaboost classifier is trained to classify each voxel of the EZ as disrupted or non-disrupted. Finally, blood vessel silhouettes and isolated points are excluded. To demonstrate its effectiveness, the proposed framework was tested on 15 eyes with retinal trauma and 15 normal eyes. For the eyes with retinal trauma, the sensitivity (SEN) was 85.69% ± 9.59%, the specificity (SPE) was 85.91% ± 5.48%, and the balanced accuracy rate (BAR) was 85.80% ± 6.16%. For the normal eyes, the SPE was 99.03% ± 0.73%, and the SEN and BAR levels were not relevant. Our framework has the potential to become a useful tool for studying retina trauma and other conditions involving EZ integrity.

Figure 1. Examples of EZ disruption region detection results and ground truths for a subject with retinal trauma.

The red region represents the ground truth, and the yellow region represents the segmented EZ disruption region using the proposed method. (a–c) Three original B-scans of an OCT volume. (d–f) The corresponding ground truth in the B-scans for (a–c), respectively. (g–i) The corresponding detection results using the proposed method for (a–c), respectively. (j) The ground truth in a 3D view. (k) The detection results in a 3D view. (l) The en face projection of the VOIs. (m) The en face projection of the ground truth (in red). (n) The en face projection of the detection results (in yellow).

Figure 2. An example of the detection results using the proposed method on a normal subject.

(a,b) The original B-scans of the OCT volume. (c,d) The false positive detection results (in green) using the proposed method. (e) All false positive detection results in a 3D view. (f) The en face projection of the VOIs. (g) The en face of the false positives (in green).

Figure 3. Detected EZ disruption volume comparison.

The blue bars show the mean volumes, and the red error bars show the 95% confidence intervals.

Figure 4. Bland-Altman plot for consistency analysis.

Figure 5. The effect of incorrect surface segmentation and poor quality in the SD-OCT image.

(a) The original B-scans of the OCT volume. (b) The incorrectly segmented 7^th and 8^th surfaces. (c) The ground truth (in red). (d) The segmented EZ disruption (in yellow). (e) The original B-scans of the OCT volume with poor quality. (f) The ground truth (in red). (g) The segmented EZ disruption (in yellow).

Figure 6. Segmentation results of 11 intra-retinal surfaces (10 layers) on a normal eye and an eye with retinal trauma.

(a) B-scan of a normal eye. (b) Segmentation results of the normal eye, nerve fibre layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer unclear layer (ONL) + inner segment layer (ISL), outer segment layer (OSL), and retinal pigment epithelium complex (RPE+ ). (c) Three-dimensional rendering of the segmented surfaces for the normal eye. (d) B-scan of an eye with retinal trauma. (e) Segmentation results of the eye with retinal trauma. (f) Three-dimensional rendering of the segmented surfaces for the eye with retinal trauma.

Figure 7. Vessel silhouettes appear as similar low-intensity regions in the EZ disruption regions.

(a) The arrows indicate vessel silhouettes in one of the slices of the retinal OCT image. (b) The red line indicates the location of the slice shown in (a) at en face project image of the retina. The locations where vessels cross the slice and the locations of the vessel silhouettes in the same slice correspond one-to-one using differently coloured arrow pairs.