Fully-Automatic 3D Intuitive Visualization of Age-Related Macular Degeneration Fluid Accumulations in OCT Cubes

Abstract

Age-related macular degeneration is the leading cause of vision loss in developed countries, and wet-type AMD requires urgent treatment and rapid diagnosis because it causes rapid irreversible vision loss. Currently, AMD diagnosis is mainly carried out using images obtained by optical coherence tomography. This diagnostic process is performed by human clinicians, so human error may occur in some cases. Therefore, fully automatic methodologies are highly desirable adding a layer of robustness to the diagnosis. In this work, a novel computer-aided diagnosis and visualization methodology is proposed for the rapid identification and visualization of wet AMD. We adapted a convolutional neural network for segmentation of a similar domain of medical images to the problem of wet AMD segmentation, taking advantage of transfer learning, which allows us to work with and exploit a reduced number of samples. We generate a 3D intuitive visualization where the existence, position and severity of the fluid were represented in a clear and intuitive way to facilitate the analysis of the clinicians. The 3D visualization is robust and accurate, obtaining satisfactory 0.949 and 0.960 Dice coefficients in the different evaluated OCT cube configurations, allowing to quickly assess the presence and extension of the fluid associated to wet AMD.

Keywords: 3D visualization; Age-related macular degeneration; Computer-aided diagnosis; Optical Coherence Tomography.

Fig. 1

On the left, the fundus image with the analysis area marked and a green line marking the particular slice. On the right, the OCT slice is shown

Fig. 2

OCT image showing subretinal fluid associated with AMD (surrounded by a yellow box)

Fig. 3

Example of OCT image (left) and corresponding mask (right)

Fig. 4

Diagram showing the methodology used to carry out the final 3D reconstruction

Fig. 5

Individual examples of each transformation on the same image

Fig. 6

Several examples of the overall application of all the transformations on the same image. The first 2 images correspond to the original image and the original mask

Fig. 7

Evolution of the training loss in red and the validation loss in blue throughout the training process

Fig. 8

Example showing network segmentation errors in three successive slices with a transition to non-fluid areas. Row one shows 3 close slices of a cube, row 2 shows the masks made by the expert for these slices and row 3 shows the prediction made by the network

Fig. 9

Representation and comparison between some images (Image columns) of the test cubes, the segmentation made by the human expert (Label columns) and the mask predicted by the network (Pred columns)

Fig. 10

3D reconstruction of the fundus of the patient. The fluid associated with AMD is marked in the color range from red to green. From left to right: cube reconstructed with the masks annotated by the expert, cube reconstructed with the masks predicted by the net and cube reconstructed with the masks predicted by the net using the smoothing

Fig. 11

Several representations of the different interactive 3D models generated with the set of volumes of the OCT256 and OCT320 configurations. In each pair of cubes the left image corresponds to the cube created using the masks provided by the expert, while the right image corresponds to the reconstruction using our model