Director: Ruikang (Ricky) Wang

GA segmentation

 

 

Purpose

To segment geographic atrophy (GA) in OCT scans, based on the hyper-transmission defects by machine learning algorithm.

 

Input

Zeiss OCT/OCTA IMG files 6x6 macular scans

 

Output

 Pixel wise 2D GA mask (.png)

 GA area with (.csv)

 manual correction 2D GA mask (.png)

 

Validation

Chu, Zhongdi, et al. "Automatic geographic atrophy segmentation using optical attenuation in OCT scans with deep learning." Biomedical Optics Express 13.3 (2022): 1328-1343.

 

 

The interface of the software

 

The 3-channel composite image was used as the input for the software. The hyperTD region was segmented by the algorithm.

The b-scan checking tool to review the lesion through OCT 3D volumes.

 

The result of automatically segmentation can be manually edited by the grader.

 

Each identical lesion was labeled after adjusted by the human grader, and its area was calculated by the software.