Quantitative analysis of the Stratus optical coherence tomography fast macular thickness map reports

The cross sectional optical coherence tomography images have an important role in evaluating retinal diseases. The reports generated by the Stratus fast macular thickness scan protocol are useful for both clinical and research purposes. The centerpoint thickness is an important outcome measure for many therapeutic trials related to macular disease. The data is susceptible to artifacts such as decentration and boundary line errors and could be potentially erroneous. An understanding of how the data is generated is essential before utilizing the data. This article describes the interpretation of the fast macular thickness map report, assessment of the quality of an optical coherence tomography image and identi ﬁ cation of the artifacts that could in ﬂ uence the numeric data.

Longitudinal assessment of the foveal thickness on OCT is an important outcome measure for many therapeutic trials. The fundus photograph reading center (FPRC), Madison WI, USA, serves as a central laboratory to receive and analyze OCTs for multicenter clinical trials. The scans are taken by trained and certifi ed operators and evaluated at the reading center by ocular disease evaluators (ODE). Quality of the OCT is one of the most important evaluation procedures used by the ODE since it determines if the soft ware-generated numeric data is erroneous or not. Table 1 summarizes the artifacts frequently seen in OCTs. Artifacts generated by the soft ware are common and could potentially produce erroneous data. [12][13][14] Manual remeasurement of foveal thickness is performed in scans with faulty centerpoint thickness (CPT). Over 30,000 OCT scans are evaluated annually at the reading center and nearly 30% need manual remeasurement.
An understanding of the soft ware algorithms and how each of the elements of the report is generated is essential for interpreting the thickness measurements. The purpose of this article is to assist with interpretation of the reports generated by the FMTM protocol and understanding of the artifacts generated by the OCT.

The fast macular thickness map protocol
The combined data from the six radial scans of the FMTM scanning protocol is represented in the FMTM retinal map analysis report. The individual component B-scans are available on retinal thickness reports. The map report should always be analyzed in conjunction with the six individual thickness reports. The map report has six components to it [ Fig. 1].
(1) The B-scan (OCT image) is a truncated image of the scan with the least signal intensity. (2) The fundus image is taken aft er the scan is complete and can show the location of the radial scan lines and the centerpoint (intersection of the six lines). (3) The signal strength is a quality measurement which employs the signal/noise ratio on a scale of 1 to 10. Signal strength tends to correlate with the contrast between layers of the retina and thus the boundary line algorithm of the machine. A signal strength less than 5 frequently indicates poor scan quality. The confi dence message given by the soft ware (available in the newer soft ware versions) informs the clinician

Amitha Domalpally, Ronald P Danis, Dawn Myers, Christina N Kruse
The cross sectional optical coherence tomography images have an important role in evaluating retinal diseases. The reports generated by the Stratus fast macular thickness scan protocol are useful for both clinical and research purposes. The centerpoint thickness is an important outcome measure for many therapeutic trials related to macular disease. The data is susceptible to artifacts such as decentration and boundary line errors and could be potentially erroneous. An understanding of how the data is generated is essential before utilizing the data. This article describes the interpretation of the fast macular thickness map report, assessment of the quality of an optical coherence tomography image and identifi cation of the artifacts that could infl uence the numeric data.
132 Alignment artifacts [17] Distortion/fl attening of morphology in AMD/central serous chorioretinopathy following automated retinal thickness (single eye) analysis on Stratus software Artifacts due to low signal strength Layers on the B-scan not clearly seen The quality of the OCT scan needs to be evaluated before interpreting any of the components. Fig. 1 demonstrates an example of a good quality OCT image. The B-scan has correctly placed boundary lines, the radial scan lines on the fundus image seem to intersect at the macula, the signal strength is good (Ͼ5). The standard deviation of the CPT is less than 10% of the CPT. The fovea (blue area) is positioned in the central subfi eld of the pseudocolor map. The numeric data available on the report of a good quality OCT can be readily accepted. Poor quality OCT scans have artifacts that aff ect their numeric data. The CPT can be remeasured using commercially available soft ware (Stratus review soft ware); however, the subfi eld values cannot be reclaimed. Table 1 gives a list of possible artifacts with the Stratus OCTs. Two major errors that could aff ect the numeric data and occur quite commonly are boundary line errors and decentration.

Boundary line errors
There are two boundary lines identifi ed by the soft ware in a cross-sectional retinal scan-the fi rst hyper-refl ective band is the internal limiting membrane (ILM) and the second hyper-refl ective band is considered to be the retinal pigment epithelium (RPE). Studies have shown that the second narrow hyper-refl ective band in normal eyes is the photoreceptor inner segment-outer segment (IS-OS) junction and that the true RPE is missed by the segmentation algorithm. [15,16] Errors in the boundary line could occur due to misidentifi cation of either the ILM or RPE. Vitreous detachment, macular holes, epiretinal membranes, and large cysts are some of the morphological abnormalities that could create boundary line errors of the ILM. The RPE boundary could be erroneously drawn in eyes with neovascular lesions, or other abnormalities which would create hyper-refl ectivity such as blood and hard exudates. Errors of both layers could occur in images with low signal strength or poor scan registration (Z off set error). Retinal thickness is calculated as the distance between the two boundary lines. Errors in the boundary lines could therefore adversely aff ect the measurements. If the error involves the centerpoint of the scan, the CPT is aff ected. In such cases, the Stratus review soft ware can be used to remeasure CPT by placing the soft ware calipers at the true boundary lines of the centerpoint [Fig. 4]. If the boundary line error aff ects any other part of the scan, then the subfi eld values of the corresponding radial scan are incorrect.
Boundary line errors can easily be identified using the six individual B-scans. Wedge or bowtie artifacts on the pseudocolor map and a high standard deviation of the CPT are additional clues to the presence of this artifact [Fig. 5]. On occasion, rescanning may avoid this artifact but mostly, these errors are unavoidable and result in lost data. The soft ware versions have been upgraded to correct boundary

Decentration
CPT is an important parameter for evaluation in macular disease because it represents foveal thickness. An important aspect to quality assessment is to confi rm whether the CPT truly represents the foveal thickness. If the radial lines of the scan patt ern have not been centered on the fovea, the scan is considered decentered and the CPT would not represent the foveal thickness. The center of the macula corresponds to A-scan 64 as shown in Fig. 2 and the central subfi eld of the numeric plot is assumed to correspond to the macula in well-centered scans as shown in Fig. 3. In decentered scans, the CPT does not represent true foveal thickness. The numeric plot is also shift ed and the values of the subfi elds do not represent their presumed location on the fundus.
Decentration artifact can be assessed using a combination of clues [ Fig. 6]. In scans with a visible foveal depression, if the fovea is not located at A-scan 64 the scan is assumed to be decentered. The OCT image on the map report is truncated and should not be used for assessing decentration. The six underlying B-scans from the retinal thickness reports can be used to evaluate decentration. The radial lines should intersect at the center of the macula in the fundus image. However, it should be kept in mind that the fundus image is taken aft er the OCT scan is completed. In thin retinas with normal foveal depression, the fovea is represented by a blue circular area on the pseudocolor map and is located within the central subfi eld if well centered. The standard deviation may or may not be high in decentered scans. If the foveal depression is not clearly visible, alternative clues are used to identify the fovea such as change in the refl ectivity of the inner layer due to loss of ganglion cell layer at the fovea, proximity of the largest cyst in cystoid macular edema (assumed to be close to the fovea), and examination of the cross hair scans which may bett er display the fovea and can be used to reference the location on the fast macular scans.
Decentration is mostly an operator-dependent error. Poor patient fi xation or inability to identify the fovea in a distorted retina could be contributory factors. However, if identifi ed, rescanning could help avoid this artifact. In decentered scans, remeasurement of foveal thickness using the soft ware calipers (Stratus review soft ware) is a possibility. The retinal thickness report with an identifi able fovea is selected and the calipers are placed at the inner and outer boundary lines of the point identifi ed as fovea. This gives the true foveal measurement. The remaining data in the subfi elds cannot be reclaimed.
In summary, artifacts on OCTs performed using the FMTM scan protocol of the Stratus OCT are common. Identifi cation of operator-dependent artifacts and retaking images if required reduces the frequency of poor quality OCTs in clinical trials. Simple measures such as centering the scan in a well-dilated pupil, focus and z off set adjustment, and having the patient close their eyes to wet the cornea before taking the image also improves image quality. Poor quality OCTs with erroneous