Abstract
Purpose :
The laser-induced choroidal neovascularization (L-CNV) murine model provides quantifiable data to evaluate efficacy of therapeutic treatments and genetic effects. We previously validated lesion quantification from optical coherence tomography (OCT) images as a reliable assessment of CNV lesion size. However, the existing method is time consuming, and the subjective nature of defining lesion borders led to variability between evaluators. We therefore investigated if OCT lesion quantification could be simplified, streamlined, and made more reliable among evaluators.
Methods :
The rate-limiting step in analysis is defining the borders of a given lesion. We developed a workflow where the lesion borders are specified using the multi-point tool in ImageJ. The lesion borders were pinpointed to the pixel coordinates in the x-plane and y-plane. These coordinates were imported to a custom Excel template and converted to µm, adjusting for the different degrees of scaling in the OCT x and y planes. Volume was calculated as previously described. This streamlined method was compared against the prior method to evaluate reliability and quantification time per lesion. To evaluate the interrater reliability, 10 OCT images of lesions were scored by 6 trained evaluators and superimposed to assess consistency.
Results :
The new process demonstrated improved time efficiency and precise quantification of OCT lesion volume. The time to calculate the volume of a lesion previously was 2 minutes, 56.9±16.4 seconds (mean±SD). The new method averaged 22.0±1.9 seconds, or ~8 times faster (paired t-test, p<0.0001, n=5). Subjectively, evaluators noted that the new method allowed for greater reliability in lesion border identification. Some lesions had near unanimous agreement with only minor differences between volume measurements while others had larger differences. The largest discrepancy between evaluators was the determination of the superior-most coordinates. Expert consensus demonstrated the importance of identifying the lesion as an ellipsoid shape and avoiding edema or a bleb as the superior coordinate of the lesion.
Conclusions :
A new method of evaluating L-CNV lesion size greatly reduced analysis time, and subjectively allowed for more precise lesion border identification. Inter-evaluator discrepancies exist but may be minimized by group discussion and training.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.