Computed tomography image processing and data analysis
View Patent ↗Systems and methods for computed tomography image processing and data analysis. An example image may comprise: receiving a plurality of input images of a scanned body; processing, by a first convolutional neural network, the plurality of input images to produce one or more pseudo radioisotope emission (pseudo-RE) images of the scanned body; and processing, by a second convolutional neural network, the plurality of pseudo-RE images to classify one or more input images of the plurality of input images with respect to a specified medical diagnosis with respect to the scanned body.
1. A method comprising:
receiving a plurality of input images of a scanned body;
processing, by a first convolutional neural network, the plurality of input images to produce one or more pseudo radioisotope emission (pseudo-RE) images of the scanned body, wherein the pseudo-RE images emulate radioactive tracer distribution within the scanned body at a cellular level;
processing, by a second convolutional neural network, the plurality of pseudo-RE images to classify one or more input images of the plurality of input images with respect to a specified medical diagnosis with respect to the scanned body; and
determining an error margin associated with classifying one or more input images, wherein the error margin reflects a ratio of true positive and false positive outcomes.
2. The method of claim 1 , wherein processing the plurality of pseudo-RE images yields a probability of one or more input images of the plurality of input images comprising diagnostically-significant image features associated with the specified medical diagnosis with respect to the scanned body.
3. The method of claim 2 , wherein the probability of the input images comprising the diagnostically-significant image features is determined by a cross-entropy function.
4. The method of claim 3 , wherein an error signal of the cross-entropy function associated with a second convolution neural network is proportional to a difference between desired and actual output values.
5. The method of claim 1 , wherein processing the plurality of pseudo-RE images yields one or more regions of interest (ROIs) within one or more input images of the plurality of input images, wherein each ROI comprises diagnostically-significant image features associated with the specified medical diagnosis with respect to the scanned body.
6. The method of claim 1 , wherein processing the plurality of pseudo-RE images yields one or more annotated images, wherein each annotated image is associated with a corresponding input image of the plurality of input images.
7. The method of claim 1 , further comprising:
processing a training data set comprising a first plurality of annotated input images and a second plurality of corresponding RE images to determine one or more parameters of at least one of: the first convolutional neural network or the second convolutional neural network.
8. The method of claim 7 , wherein processing the training data set further comprises at least one of: cropping one or more input images of the plurality of input images or cropping one or more RE images of the plurality of RE images.
9. The method of claim 7 , wherein a first pixel resolution of an annotated input image of the plurality of annotated input images exceeds a second pixel resolution of a corresponding RE image.
10. The method of claim 1 , wherein a first pixel resolution of an input image of the plurality of input images exceeds a second pixel resolution of a pseudo-RE image of the plurality of pseudo-RE images.
11. The method of claim 1 , wherein the plurality of input images comprises one or more computed tomography (CT) images.
12. The method of claim 1 , wherein the plurality of pseudo-RE images emulate Positron Emission Tomography (PET) images.
13. A system, comprising:
a memory;
a processor, coupled to the memory, the processor configured to:
receive a plurality of input images of a scanned body;
process, by a first convolutional neural network, a plurality of input images of a scanned body to produce one or more pseudo radioisotope emission (pseudo-RE) images of the scanned body, wherein the pseudo-RE images emulate radioactive tracer distribution within the scanned body at a cellular level;
process, by a second convolutional neural network, the plurality of pseudo-RE images to classify one or more input images of the plurality of input images with respect to a specified medical diagnosis with respect to the scanned body; and
determine an error margin associated with classifying one or more input images, wherein the error margin reflects a ratio of true positive and false positive outcomes.
14. The system of claim 13 , wherein processing the plurality of pseudo-RE images yields a probability of one or more input images of the plurality of input images comprising diagnostically-significant image features associated with the specified medical diagnosis with respect to the scanned body.
15. The system of claim 13 , wherein processing the plurality of pseudo-RE images yields one or more regions of interest (ROIs) within one or more input images of the plurality of input images, wherein each ROI comprises diagnostically-significant image features associated with the specified medical with respect to the scanned body.
16. The system of claim 13 , wherein the processor is further configured to:
processing a training data set comprising a first plurality of annotated input images and a second plurality of corresponding RE images to determine one or more parameters of at least one of: the first convolutional neural network or the second convolutional neural network.
17. The method of claim 16 , wherein processing the training data set further comprises at least one of: cropping one or more input images of the plurality of input images or cropping one or more RE images of the plurality of RE images.
18. The system of claim 13 , wherein a first pixel resolution of an input image of the plurality of input images exceeds a second pixel resolution of a pseudo-RE image of the plurality of pseudo-RE images.
19. The system of claim 13 , wherein the plurality of input images comprises one or more computed tomography (CT) images.
20. The system of claim 13 , wherein the plurality of pseudo-RE images emulate Positron Emission Tomography (PET) images.
21. A computer-readable non-transitory storage medium comprising executable instructions that, when executed by a computing device, cause the computing device to perform operations comprising:
receiving a plurality of input images of a scanned body;
processing, by a first convolutional neural network, a plurality of input images of a scanned body to produce one or more pseudo radioisotope emission (pseudo-RE) images of the scanned body, wherein the pseudo-RE images emulate radioactive tracer distribution within the scanned body at a cellular level; and
processing, by a second convolutional neural network, the plurality of pseudo-RE images to yield a probability of one or more input images of the plurality of input images comprising diagnostically-significant image features associated with a specified medical diagnosis with respect to the scanned body, wherein the probability of the input images comprising the diagnostically-significant image features is determined by a cross-entropy function, and wherein an error signal of the cross-entropy function network is proportional to a difference between expected and actual output values.
22. The computer-readable non-transitory storage medium of claim 21 , wherein processing the plurality of pseudo-RE images yields a probability of one or more input images of the plurality of input images comprising diagnostically-significant image features associated with the specified medical diagnosis with respect to the scanned body.