SD-OCT Retinal Image Analysis

SD-OCT images were manually analyzed with the caliper feature on the Heidelberg Spectralis commercial software by one of the authors (KK) and validated by the senior author (AHK). No commercially available, automated measurements of the retina were used. The image contrast was manually set for each individual SD-OCT scan to optimize the signal-to-noise ratio and allow identification of relevant retinal anatomy, especially the retinal layers. Manual measurements were made to quantify the inner retinal thickness (IRT), outer retinal thickness (ORT), and full retinal thickness (FRT) in the 1:1-μm display, including those for oblique measurements. Although it was possible to identify the approximate location of outer retinal layers in subjects with SMH, the identification of the outer boundary of the ONL was at the discretion of the graders' best judgment and possibly prone to error. Therefore, we performed separate manual measurements of the IRT and ORT to remove any potential bias that would be introduced by manual measurements of the ORT. The IRT included the internal limiting membrane (ILM), retinal nerve fiber layer, ganglion cell layer, inner plexiform layer, and inner nuclear layer (INL). The ORT included the outer plexiform layer (OPL) and ONL. The border of the IRT and ORT at the junction between the INL and OPL was chosen, because this location was observed to be most clearly delineated across all pre- and postoperative SD-OCT scans (making the measurements most consistent and accurate for the IRT and ORT). The FRT spanned from the ILM to the outer boundary of the ONL.

Although a published nomenclature²⁴ outlined the retinal landmarks, and another previously published study²⁵ concluded that the automated Heidelberg Spectralis SD-OCT defines the full thickness retina to the outer boundary of the RPE, our dataset was limited by the variability in retinal anatomy and obscuration of the RPE, especially in preoperative SD-OCT scans of the retina overlying the SMH. To obtain consistent results across all scans, we avoided automated measurements and performed manual measurements in the areas overlying the SMH, corresponding normal areas in the untreated contralateral eye, and normal areas immediately adjacent to the SMH in the treated eye. However, measurements of the neurosensory retina spanning from the ILM to the inner boundary of the RPE were also performed in the normal, adjacent areas to detect changes in significant differences. In five pre- and one postoperative SD-OCT scans (Figs. 1A, ​A,1B),1B), the retina was oriented obliquely relative to the x-axis, so measurements were made perpendicular to the axis of the retina, rather than parallel to the green OCT line marker. The locations of the outer retinal landmarks were manually interpolated from adjacent normal regions in cases where the boundary of the outer layers was difficult to directly visualize due to the adjacent subretinal hemorrhage. This method has been previously described.²⁶ The visibility of the ELM, EZ, and RPE were also noted on each scan that was used for measurements.

Fundus photographs and representative SD-OCT images of patient 3. (A) SD-OCT section of SMH involving the fovea prior to surgery shows a thickened inner and outer retina. The RPE is not visible underneath the SMH. The red line represents the measurement made for the IRT, the yellow line represents the ORT, and the green line represents the FRT. (B) Postoperative month 11.5 SD-OCT scan of the same area shows resolution of the SMH. The IRT and ORT appear qualitatively thinner. The ELM and RPE are also apparent now. The same color scheme in A is used to highlight the measurements for the retinal thicknesses. (C) Infrared fundus image of the retina with the bold green line representing the location of SD-OCT retinal thickness measurements before and after subretinal tPA displacement of SMH. (D) Preoperative fundus photograph shows the SMH covering the fovea. (E) Postoperative month 1 fundus photograph shows resolution of the SMH.

Retinal thickness measurements of the detached area overlying the SMH were obtained prior to surgery and repeated in the same areas following surgery. Many subjects did not have pre- or postoperative imaging sufficient for inclusion in this study. To perform standardized and reliable measurements, all measurements were obtained within a predefined circular annulus that had an approximate inner and outer diameter of 3 and 4 mm, respectively, and was centered on the fovea (Fig. 2). This predefined region of interest (ROI) was selected such that it would maximize the possibility of including both detached retina as well as normal retina within its boundaries across all cases. By performing retinal measurements in normal and abnormal areas within this ROI, the retinal thickness overlying areas of SMH could be compared with retinal thickness measurements from the contralateral eye but symmetrical locations of the macula that served as controls. In this way, the normal variations in retinal thickness due to retinal topology was not a confounding factor in the measurements.

Fundus photographs and representative SD-OCT images of patient 6. (A) SD-OCT scan of a SMH involving the fovea prior to surgery shows thickened inner and outer retina. The ELM and EZ are not apparent in the area of the hemorrhage. The blue rectangular region adjacent to the hemorrhage highlights a normal area unaffected by the SMH but later detached by the induced RD. The red line represents the IRT, the yellow line represents the ORT, and the green line represents the FRT. (B) Postoperative month 11 SD-OCT scan of the same area captured in previous panel shows resolution of the SMH. The high-magnification image shows reappearance of the ELM and EZ. The same color scheme in A is used to highlight the measurements for the retinal thicknesses. (C) Infrared fundus image of macula with SD-OCT B-scan locations denoted in green lines. The two concentric yellow circles centered on the fovea with diameters of approximately 3 and 4 mm, respectively, are shown. The bold green line represents the location within the concentric circles where the SD-OCT–based retinal thickness measurements were made before and after the subretinal tPA displacement of SMH. (D) Preoperative fundus photograph showing the SMH. (E) Postoperative fundus photograph at month 2.5 of the nearly resolved SMH depicted in previous panel.

In most cases, measurements of the retinal thicknesses were made at the 12 and 6 o'clock locations on the ROI for consistency and comparability between individual measurements on different dates and among subjects. If a normal region of attached retina was not observed within the prespecified ROI in the pathologic eye, similar corresponding measurements were made preoperatively in the contralateral unaffected eye to serve as controls. Figure 3 is another schematic of preoperative and postoperative fundus photography illustrating the locations where retinal thickness measurements were made in the areas overlying the SMH and, if possible, in the normal unaffected locations immediately adjacent to the SMH. In addition, the retina highlighted in the blue box of Figure 2A is an example of a normal region that was unaffected by the SMH but affected by the induced focal, short-term retinal detachment. If normal areas were available for assessment on the OCT scans, qualitative assessments of the RPE, ELM, and EZ were also done in these areas.

Schematic of the fundus locations where OCT measurement were made before and after subretinal tPA displacement of SMH. (A–C) Fundus photographs demonstrate large SMH. (D–F) Photographs demonstrate resolution of the hemorrhages following surgery and SMH displacement with subretinal tPA at 13, 11, and 1 months, respectively. The blue points represent the location of OCT retinal thickness measurements overlying the hemorrhage, and the yellow points represent the location of measurements made immediately adjacent to the hemorrhage (attached retina). Normal, attached retinal OCT sections were not available in regions immediately adjacent to the SMH for the eyes in A and D, so yellow points are not depicted.