Computerized device and method for processing image data
A computerized device for processing image data is proposed. The computerized device comprises a receiving unit which is configured to receive optical coherence tomography data of a of a tissue, in particular of a retina, a providing unit which is configured to provide a convolutional neural network for processing the optical coherence tomography data, and a processing unit which is configured to process the received optical coherence tomography data using the convolutional neural network for identifying at least one certain object in the tissue.
1. A computerized device for processing image data, the computerized device comprising:
one or more processors configured to execute instructions to:
receive optical coherence tomography data of a tissue, in particular of a retina,
provide a convolutional neural network for processing the optical coherence tomography data,
directly process raw image data included within the received optical coherence tomography data using the convolutional neural network for identifying intraretinal cystoid fluid and subretinal fluid in the received optical coherence tomography data, and
provide the convolutional neural network with a plurality of layer stacks, wherein the plurality of layer stacks simultaneously uses different two-dimensional or three-dimensional image patches of different sizes.
2. The computerized device of claim 1 ,
wherein the one or more processors are further configured to execute instructions to segment and to classify the intraretinal cystoid fluid and the subretinal fluid in the received optical coherence tomography data.
3. The computerized device of claim 2 , wherein the intraretinal cystoid fluid and the subretinal fluid are classified at least in part based on a relative position to one or more retinal layers.
4. The computerized device of claim 1 ,
wherein the one or more processors are further configured to execute instructions to localize each of the intraretinal cystoid fluids and to localize each of the subretinal fluids in the optical coherence tomography data.
5. The computerized device of claim 1 ,
wherein the one or more processors are further configured to execute instructions to quantify a volume of each of the intraretinal cystoid fluids and to quantify a volume of each of the subretinal fluids in the optical coherence tomography data.
6. The computerized device of claim 1 ,
wherein the optical coherence tomography data are spectral-domain optical coherence tomography images.
7. The computerized device of claim 1 ,
wherein each of the layer stacks includes one or more pairs of convolutional layers and max-pooling layers, one or more fully-connected layers and a terminal classification layer.
8. The computerized device of claim 7 ,
wherein the one or more processors are further configured to execute instructions to train the convolutional neural network using a number of two-dimensional image patches or a number of three-dimensional image patches extracted from a plurality of spectral-domain optical coherence tomography images.
9. The computerized device of claim 8 ,
wherein the one or more processors are further configured to execute instructions to provide the convolutional neural network, in particular the terminal classification layer, with at least one location information of the two-dimensional or the three-dimensional image patches.
10. The computerized device of claim 8 ,
wherein the one or more processors are further configured to execute instructions to provide the convolutional neural network, in particular the terminal classification layer, with Euclidean distances of patch centers of the two-dimensional or the three-dimensional image patches and three-dimensional coordinates of the patch centers.
11. The computerized device of claim 8 , wherein the convolutional neural network is trained exclusively using image patches extracted from the plurality of spectral-domain optical coherence tomography images.
12. The computerized device of claim 1 ,
wherein the different two-dimensional or the three-dimensional image patches are centered at a same position.
13. A method for processing image data via a computerized device, the method comprising:
receiving optical coherence tomography data of a tissue, in particular of a retina,
providing a convolutional neural network for processing the optical coherence tomography data,
directly processing raw image data included within the received optical coherence tomography data using the convolutional neural network for identifying intraretinal cystoid fluid and subretinal fluid in the received optical coherence tomography data, and
provide the convolutional neural network with a plurality of layer stacks, wherein the plurality of layer stacks simultaneously uses different two-dimensional or three-dimensional image patches of different sizes.