Deep Learning-based Choroidal Structural Assessment Program (DCAP) is an algorithm developed by Prof. Mingguang He and his research group, aiming at quantifying choroidal parameters on optical coherence tomography (OCT) images accurately, repeatedly, and automatically. DCAP was computed by segmenting, denoising, and binarizing OCT images using deep learning neural networks. It has been validated in children with low to moderate myopia in China, and the validation in diseased status is ongoing. More detailed information about DCAP can be found in our previous studies.


Take a try

Calculating the luminal and interstation regions of the choroid.
750um either side of the fovea center along the choroid tangent is indicated.


Image Results to save
Raw Images
Binary Images
Segmentation Predictions
Overlay Images (predicted Choroid over raw B-scan)

Denoising Options:
New Binarization Method

Note: these are applied to raw image before the thresholding (Niblack) is done
** Set Min brightness ON works well but needs more experimentation
Dehaze - dark channel prior
   ** ie 0.8 reduces img to 80% of its original size for analysis
Set Min brightness - using histogram mode of the dark values
Basic Image Smoothing
Segmentation & Volume Analysis

Selected file: (none)



Download Results!

Results of Choroidal Structure Analysis within the Region of Interest

Overlay of Binarized Choroid onto Bscan


Prof.Mingguang He
Dr.Wei Wang
Dr.Danli Shi
Dr.Meng Xuan

Inquiry for collaboration

We intend to provide DCAP as a research and educational tool free of charge for all researchers so long as the program is appropriately acknowledged. Should any researchers be interested in the collaboration, or other comments or feedback, please contact: mingguanghe@gmail.com


This website is for research purposes solely. DCAP researchers and developers do not hold any responsibility for any consequence whatsoever of its use.