Name

670 - Analyse automatisée des gouttes : une nouvelle échelle pour la classification de la dystrophie endothéliale de Fuchs

Fuchs endothelial corneal dystrophy (FECD) is a disease characterized by central deposits of corneal guttata and endothelial dysfunction, with relative sparing of the corneal peripheries. The severity of the disease is currently clinicaly grading, the most common used scale being the Krachmer classification. In this study, we created an automated FECD grading system by taking into consideration a larger set of quantitative and qualitative measures including guttata size, inter-guttata distances, guttata density and guttata distribution pattern. 

An automated image analysis algorithm was designed to facilitate acquisition of data relevant to the newly proposed FECD grading system, with the dual aims of achieving a) rapid image analysis and b) reproducibility of measurements and severity grading outcomes. A total of 40 patients who were on follow-up at the Singapore National Eye Centre for FECD were recruited between April to September 2015. Imaging was performed with each patient positioned on a slit lamp, using a flash-free digital camera capable of aquiring images at a resolution of 16 megapixels. An automated image processing algorithm was custom-built using a combination of software codes written in both Matlab and ImageJ. Each retroilluminated corneal image was divided into 4 optical zones for analysis: a) Central 4mm zone, b) Central 7mm zone c) Peripheral 4-7mm zone, and d) Enhanced Strip Circle (ESC) zone. The ESC was defined as a circle centered on the region of the cornea with greatest guttata density, which extended to totally incorporate the central 4mm zone, with the minimum possible diameter. The automated image processing software evaluated each zone in each retroilluminated corneal image to determine the total guttata count, guttata density, individual guttata sizes and inter-guttata distances. 

Mean patient age was 69.4±10.3 years old. 75% (n=30) of patients were females. A total of 86,553 guttata were measured in 40 eyes, with mean gutta size being 407 ± 155 µm2  and mean inter-guttata distance being 128±38µm. A mean total of 2163±984 guttata were identified within the 7mm zone. Both the 4mm zone and ESC zones contained significantly fewer guttata (n=968±984 and 1425±780 respectively) compared to the 7mm zone (p<0.001). However, the ESC zone included a much larger number of guttata compared to the 4mm zone, representing an average of 65.9% of all guttata contained within the 7mm zone, which was significantly greater than the 44.8% represented by the 4mm zone (p<0.001). Guttata were most commonly located along the interpalpebral region, with a slight inferio-temporal paracentral shift. Both the 4mm zone and ESC zones contained significantly fewer guttata (n=968±984 and 1425±780 respectively) compared to the 7mm zone (p<0.001). The automated image processing software was validated to detect retroilluminated guttata at a sensitivity of 95.4% (95% CI: 93.6%-97.5%), with a PPV of 94.4% (95% CI: 91.1%-97.6%). 

Instead of categorically grading FECD based on arbitrary physical parameters, our software was able to describe and grade FECD more objectively and comprehensively by taking into consideration a much larger spectrum of quantitative data. 

The automated FECD grading system will be useful for the management of FECD: differentiating patients who only need descemethorexis without endothelial keratoplasty from those who need an endothelial graft.