Abstract
Abstract: :
Purpose:To evaluate the sensitivity and specificity of a new GDx screening program, whose purpose is to detect nerve fiber layer loss associated with glaucoma and compare it to results from the full exam done with the GDx VCC. Methods: 68 eyes were analyzed from 35 subjects. These subjects were broken down into the following groups: Normals, Glaucoma Suspects based upon large cupping (> 0.6) with full 24–2 SITA visual field or Glaucoma (optic nerve damage and visual field loss). Screening and Full Examinations were performed with the GDx. The output of the screening method is Within Normal Limits, Borderline, or Outside Normal Limits based upon results of the NFI, and TSNIT average, Superior Average, and Inferior Average parameters. The results from the Full Exam were analyzed in a similar manner and the outcome of both methods was compared for agreement. Results: Agreement between screening and Full Exam was 68%. If you consider the agreement separately for each diagnostic group, glaucomatous eyes was 72%, 82% for normal eyes, and 54% for suspect eyes. For suspect eyes only, the Full Exam method classified 57% as normal, while the screening method classified 54% as normal. Diagnostic accuracy – The sensitivity and specificity for the Full Exam method were 72% and 77% respectively. The sensitivity and specificity for the screening method were 67% and 91% respectively. The suspects were not included in this calculation. The Full Exam was more sensitive, however the screening exam was more specific. Discussion The overall agreement between the Full Exam method and the screening method was relatively poor (68%). However the agreement was better for the normal eyes (82%) and glaucomatous eyes (72%), than for the suspect eyes (54%). It was expected that the agreement for the suspect eyes would be worse than for normals and glaucoma for several reasons. Normal variability between scans can affect agreement in eyes near borderline. For eyes near the border between normal and glaucoma, a few microns can be the difference between a normal and abnormal scan. The second is the database used for screening was a VCC database which may not be appropriate for eyes that had a different software compensation (ECC method) that is used as part of the screening algorithm. In conclusion, the agreement between the Screening and Full exam was around 70% using three outcome categories. When only two outcome categories are used, the agreement improves to around 80%. The screening method is more conservative than the Full Exam method and may miss some patients the Full Exam would identify, however the number of false positives for the Screening Method would be less than the Full Exam method.
Keywords: imaging/image analysis: clinical • nerve fiber layer • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)