Three dimensional optical coherence tomography imaging: advantages and advances

Abstract

Three dimensional (3D) ophthalmic imaging using optical coherence tomography (OCT) has revolutionized assessment of the eye, the retina in particular. Recent technological improvements have made the acquisition of 3D-OCT datasets feasible. However, while volumetric data can improve disease diagnosis and follow-up, novel image analysis techniques are now necessary in order to process the dense 3D-OCT dataset. Fundamental software improvements include methods for correcting subject eye motion, segmenting structures or volumes of interest, extracting relevant data post hoc and signal averaging to improve delineation of retinal layers. In addition, innovative methods for image display, such as C-mode sectioning, provide a unique viewing perspective and may improve interpretation of OCT images of pathologic structures. While all of these methods are being developed, most remain in an immature state. This review describes the current status of 3D-OCT scanning and interpretation, and discusses the need for standardization of clinical protocols as well as the potential benefits of 3D-OCT scanning that could come when software methods for fully exploiting these rich datasets are available clinically. The implications of new image analysis approaches include improved reproducibility of measurements garnered from 3D-OCT, which may then help improve disease discrimination and progression detection. In addition, 3D-OCT offers the potential for preoperative surgical planning and intraoperative surgical guidance.

Figure 1

(left) Cross-sectional B-scan through the optic nerve head acquired using SD-OCT, with the white dashed line denoting the location of the single A-scan (right) intensity profile

Figure 2

(A) Location of TD-OCT peripapillary scan shown on fundus photograph (left), and single cross-section OCT B-scan (right). (B) Location of TD-OCT radial macular scan shown on fundus photograph, with six corresponding OCT B-scans and their respective locations

Figure 3

Three-dimensional data cube of the optic nerve head region acquired using SD-OCT, showing Cartesian coordinates of scan location for reference. While this image consisted of 200×200 A-scans, the black lines represent the idea of voxels. The OCT enface image is created by summing individual A-scans.

Figure 4

OCT fundus image showing uneven signal strength (dark patches scattered throughout image) and large eye movements (red arrows)

Figure 5

OCT fundus image generated from TD-OCT volumetric data (256×256 A-scans). The scan took almost 3 minutes to acquire, and there are several blinking artifacts (black lines) as well as eye movements.

Figure 6

Vertical (left) and horizontal (right) OCT B-scans. The horizontal B-scan is taken along the fast scanning axis, while the vertical B-scan is resampled after the 3D volume has been acquired. The blue arrows point out the waviness that is attributed to axial eye movements that occur while the vertical section is acquired.

Figure 7

Original 3D-OCT enface image of healthy subject (top), C-mode slice taken after aligning each frame of the 3D-OCT image to the ILM (bottom). The red lines indicate the location of the corresponding cross-sectional scan; the three blue lines indicate the depth of the C-mode section.

Figure 8

Original 3D-OCT enface image of subject with vitreomacular traction (top), C-mode slice taken after aligning each frame of the 3D-OCT image to the ILM (bottom). The red lines indicate the location of the corresponding cross-sectional scan; the three blue lines indicate the depth of the C-mode section.

Figure 9

Original 3D-OCT enface image of patient with macular hole (top), C-mode slice taken after aligning each frame of the 3D-OCT image to the RPE (bottom). The red lines indicate the location of the corresponding cross-sectional scan; the three blue lines indicate the depth of the C-mode section.

Figure 10

Original 3D-OCT enface image of patient a disciform scar as a result of a wet age-related macular degeneration lesion (top), C-mode slice taken after aligning each frame of the 3D-OCT image to the RPE (bottom). The red lines indicate the location of the corresponding cross-sectional scan; the three blue lines indicate the depth of the C-mode section.

Figure 11

Longitudinal assessment available on the commercial TD-OCT software: a plot of RNFL thickness profiles from successive visits (top), and mean RNFL thickness values (bottom). This does not take into account operator-dependent variability of scan placement.

Figure 12

Registration of 3D-OCT (red) image to SLO image (green) using blood vessel location. Left: before registration. Right: after registration. Yellow regions are areas of overlap between the images.

Figure 13

OCT fundus image (left), RNFL thickness map (center) and individual frame showing segmentation of the RNFL (solid blue line and solid white line) on one frame (right).

Figure 14

RNFL thickness maps from the right eye of a healthy subject (left) and subject with glaucoma (right)

Figure 15

OCT fundus image (left) and RNFL thickness map (right) from a glaucoma subject with a localized inferotemporal defect (arrows)

Figure 16

OCT fundus image (left) and macular thickness map (right) from a healthy subject

Figure 17

Macular thickness maps (IRC segmentation) from the right eye of a healthy subject (left) and subject with glaucoma (right)

Figure 18

Examples of GCC thickness and significance maps in healthy and glaucoma subjects. The GCC consists of cell bodies, axons and dendrites of retinal ganglion cells.

Figure 19

Resampling a single 3D data volume. The top frame shows a vertical cross-section through the optic nerve, while the second frame shows a horizontal section through the optic nerve. The horizontal section is acquired along the fast axis of scanning, while the vertical section is acquired in the slow direction. The third frame shows an oblique section taken at the location of a nerve fiber layer defect. The bottom frame shows a resampled peripapillary scan with a diameter of 3.4 mm. Blue arrows indicate eye motion from scanning in the slow axis, and yellow arrows indicate the location of the nerve fiber layer defect.

Figure 20

Peripapillary scan extracted from 3D volume by manually marking the boundary of the optic nerve head, calculating the geometric center and resampling at a diameter of 3.4 mm.

Figure 21

Enface image showing automated detection of the ONH

Figure 22

Single OCT B-scan (left) and composite B-scan image generated by averaging 9 frames (right)

Figure 23

C-mode section perpendicular to the direction of scanning; this section appears distorted because several layers are sliced through. The red line indicates the location of the cross-sectional scan; the three blue lines indicate the depth of the C-mode section.

Figure 24

A high-density (3 × 3 mm; 300 × 300 A-scans) 3D-OCT image of the optic nerve can be used to create C-mode sections of the lamina cribrosa. The black line indicates the location of the cross-sectional scan; the three white lines indicate the depth of the C-mode section.

Figure 25

A high-density (800 × 800 μm, 300 × 300 A-scans) 3D-OCT image of the retina can be used to create C-mode sections of photoreceptors. The horizontal red line indicates the location of the cross-sectional scan; the horizontal blue lines indicate the depth of the C-mode section.

Figure 26

Summary of method for backward compatibility between TD-OCT and 3D-OCT: scan location matching. (A, B) TD-OCT Fundus video image and 3.4-mm circular cross-sectional OCT B-scan; (C, D) 3D-OCT fundus image and virtually resampled 3.4-mm circular B-scan; (E) Similarity map created from the correlation between TD-OCT data and data virtually sampled centered at each pixel within the sampling center boundary (square; color range represents correlation coefficient 0, dark blue, to 1, white); (F) 3D-OCT virtually resampled B-scan, after searching and matching the TD-OCT scan location. The 3D-OCT vessel shadows (dashed lines) match the TD-OCT vessel locations.