Systems and methods for analyzing perfusion-weighted medical imaging using deep neural networks

1. A computer-implemented method for analyzing perfusion-weighted imaging, the method comprising:

a) receiving perfusion-weighted imaging data acquired from a subject having received an injection of a contrast agent using a magnetic resonance (“MR”) imaging system;

b) modeling at least one voxel associated with the perfusion-weighted imaging data using a four-dimensional (“4D”) convolutional neural network;

c) extracting spatio-temporal features for each modeled voxel;

d) estimating at least one perfusion parameter for each modeled voxel based on the extracted spatio-temporal features; and

e) generating a report using the at least one perfusion parameter indicating perfusion in the subject.

2. The method of claim 1 , wherein the perfusion-weighted imaging data comprises at least one of dynamic susceptibility contrast (“DSC”) imaging data and dynamic contrast enhanced (“DCE”) imaging data.

3. The method of claim 1 , wherein the 4D convolutional neural network comprises spatial information and temporal information.

4. The method of claim 1 , wherein the 4D convolution neural network comprises at least one of a convolutional, a pooling, a non-linear, and a fully-connected layer.

5. The method of claim 4 , wherein the 4D convolution neural network further comprises a non-linear layer inserted between each of the convolutional layer and the pooling layer.

6. The method of claim 1 , wherein the method further comprises generating at least one input patch defined by a slice number, a width, a height, and a volume number for each modeled voxel.

7. The method of claim 6 , wherein the method further comprises providing the at least one input patch to the 4D convolutional neural network to extract the spatio-temporal features.

8. The method of claim 1 , wherein the method further comprises computing, using a classifier layer in the 4D convolution neural network, at least one probability of infarction based on the spatio-temporal features extracted.

9. The method of claim 8 , wherein the method further comprises generating a probability map using the at least one probability computed.

10. The method of claim 1 , wherein the method further comprises determining a condition of the subject based on the at least one perfusion parameter.

11. The method of claim 1 , wherein the at least one perfusion parameter comprises at least one of a blood volume (“BV”), a blood flow (“BF”), a mean transit time (“MTT”), a time-to-peak (“TTP”), a time-to-maximum (“T max ”), a maximum signal reduction (“MSR”), and a first moment (“FM”).

12. A system for analyzing perfusion-weighted imaging, the system comprising:

an input configured to receive perfusion-weighted imaging data acquired from a subject using a magnetic resonance (“MR”) imaging system;

a processor programmed execute instructions stored in a non-transitory computer readable media to:

i) access the perfusion-weighted imaging data;

ii) generate an input patch for each selected voxel associated with the perfusion-weighed imaging data;

iii) apply a four-dimensional (“4D”) convolutional neural network to the generated input patch to extract spatio-temporal features;

iv) compute, at least one perfusion parameter using the spatio-temporal features extracted;

v) generate a report using the at least one perfusion parameter; and

an output for displaying the report to communicate perfusion in the subject.

13. The system of claim 12 , wherein the perfusion-weighted imaging data comprises at least one of dynamic susceptibility contrast (“DSC”) imaging data and dynamic contrast enhanced (“DCE”) imaging data.

14. The system of claim 12 , wherein the 4D convolutional neural network comprises spatial information and temporal information.

15. The system of claim 12 , wherein the 4D convolution neural network comprises at least one of a convolutional, a pooling, a non-linear, and a fully-connected layer.

16. The system of claim 15 , wherein the 4D convolution neural network further comprises a non-linear layer is inserted between each of the convolutional layer and the pooling layer.

17. The system of claim 12 , wherein the processor is further programmed to compute, using a classifier layer in the 4D convolution neural network, at least one probability of infarction based on the spatio-temporal features extracted.

18. The system of claim 17 , wherein the processor is further programmed to generate a probability map using the at least one probability computed for each of plurality of selected voxels.

19. The system of claim 12 , wherein the at least one perfusion parameter comprises at least one of a blood volume (“BV”), a blood flow (“BF”), a mean transit time (“MTT”), a time-to-peak (“TTP”), a time-to-maximum (“T max ”), a maximum signal reduction (“MSR”), and a first moment (“FM”).

20. The system of claim 12 , wherein the processor is further programmed to generate at least one image by the at least one perfusion parameter computed for each of a plurality of selected voxels.

21. The system of claim 11 , wherein the processor is further programmed to determine a condition of the subject based on the at least one perfusion parameter.

22. The method of claim 1 , wherein the 4D convolutional neural network is configured to capture spatial for each modeled voxel in three dimensions (“3D”) and capture temporal information of each modeled voxel in a fourth dimension of the 4D convolutional neural network.

23. The system of claim 12 , wherein the 4D convolutional neural network is configured to capture spatial for each modeled voxel in three dimensions (“3D”) and capture temporal information of each modeled voxel in a fourth dimension of the 4D convolutional neural network.