Domain adaption for prostate cancer detection
Systems and methods for performing a medical imaging analysis task using a machine learning based model are provided. One or more input medical images acquired using one or more out-of-distribution image acquisition parameters and having out-of-distribution imaging properties are received. The one or more out-of-distribution image acquisition parameters and the out-of-distribution imaging properties are out-of-distribution with respect to training data on which the machine learning based model is trained. One or more synthesized medical images are generated from the one or more input medical images using a machine learning based generator network. The one or more synthesized medical images are generated for one or more in-distribution image acquisition parameters and have in-distribution imaging properties. The one or more in-distribution image acquisition parameters and the in-distribution imaging properties are in-distribution with respect to the training data on which the machine learning based model is trained. The medical imaging analysis task is performed based on the one or more synthesized medical images using the machine learning based model. Results of the medical imaging analysis task are output.
1 . A computer-implemented method for performing a medical imaging analysis task using a machine learning based model, comprising:
receiving one or more input medical images acquired using one or more out-of-distribution image acquisition parameters and having out-of-distribution imaging properties, the one or more out-of-distribution image acquisition parameters and the out-of-distribution imaging properties being out-of-distribution with respect to training data on which the machine learning based model is trained;
generating one or more synthesized medical images from the one or more input medical images using a machine learning based generator network, the one or more synthesized medical images being generated for one or more in-distribution image acquisition parameters and having in-distribution imaging properties, the one or more in-distribution image acquisition parameters and the in-distribution imaging properties being in-distribution with respect to the training data on which the machine learning based model is trained, the machine learning based generator network comprising 1) an encoder network, conditioned based on values of at least one of the one or more out-of-distribution image acquisition parameters, for encoding the one or more input medical images into feature representations and 2) a decoder network, conditioned based on values of at least one of the one or more in-distribution image acquisition parameters, for decoding the feature representations to generate the one or more synthesized medical images;
performing the medical imaging analysis task based on the one or more synthesized medical images using the machine learning based model; and
outputting results of the medical imaging analysis task.
2 . The computer-implemented method of claim 1 , wherein the out-of-distribution image acquisition parameters and the in-distribution image acquisition parameters comprise parameters of an MRI (magnetic resonance imaging) scanner.
3 . The computer-implemented method of claim 2 , wherein the parameters of the MRI scanner comprise b-value settings.
4 . The computer-implemented method of claim 1 , wherein the machine learning based generator network is jointly trained with another machine learning based generator network, the other machine learning based generator network generating synthesized out-of-distribution medical images from in-distribution training medical images.
5 . The computer-implemented method of claim 4 , wherein the machine learning based generator network generates reconstructed in-distribution medical images from the synthesized out-of-distribution medical images and wherein the machine learning based generator network is trained based on segmentation predictions from the reconstructed in-distribution medical images generated using the machine learning based model.
6 . The computer-implemented method of claim 1 , wherein the machine learning based generator network is trained using a discriminator network, the discriminator network distinguishing between 1) synthesized in-distribution images generated by the machine learning based generator network and corresponding segmentation predictions generated from the machine learning based model and 2) real in-distribution images and corresponding segmentation predictions as being real or synthesized.
7 . The computer-implemented method of claim 1 , wherein the one or more input medical images comprises one or more mpMRI (multi-parametric magnetic resonance imaging) images.
8 . The computer-implemented method of claim 1 , wherein the one or more input medical images depict a prostate of a patient and the medical imaging analysis task is detection of prostate cancer.
9 . An apparatus for performing a medical imaging analysis task using a machine learning based model, comprising:
means for receiving one or more input medical images acquired using one or more out-of-distribution image acquisition parameters and having out-of-distribution imaging properties, the one or more out-of-distribution image acquisition parameters and the out-of-distribution imaging properties being out-of-distribution with respect to training data on which the machine learning based model is trained;
means for generating one or more synthesized medical images from the one or more input medical images using a machine learning based generator network, the one or more synthesized medical images being generated for one or more in-distribution image acquisition parameters and having in-distribution imaging properties, the one or more in-distribution image acquisition parameters and the in-distribution imaging properties being in-distribution with respect to the training data on which the machine learning based model is trained, the machine learning based generator network comprising 1) an encoder network, conditioned based on values of at least one of the one or more out-of-distribution image acquisition parameters, for encoding the one or more input medical images into feature representations and 2) a decoder network, conditioned based on values of at least one of the one or more in-distribution image acquisition parameters, for decoding the feature representations to generate the one or more synthesized medical images;
means for performing the medical imaging analysis task based on the one or more synthesized medical images using the machine learning based model; and
means for outputting results of the medical imaging analysis task.
10 . The apparatus of claim 9 , wherein the out-of-distribution image acquisition parameters and the in-distribution image acquisition parameters comprise parameters of an MRI (magnetic resonance imaging) scanner.
11 . The apparatus of claim 10 , wherein the parameters of the MRI scanner comprise b-value settings.
12 . The apparatus of claim 9 , wherein the machine learning based generator network is trained using a discriminator network, the discriminator network distinguishing between 1) synthesized in-distribution images generated by the machine learning based generator network and corresponding segmentation predictions generated from the machine learning based model and 2) real in-distribution images and corresponding segmentation predictions as being real or synthesized.
13 . A non-transitory computer readable medium storing computer program instructions for performing a medical imaging analysis task using a machine learning based model, the computer program instructions when executed by a processor cause the processor to perform operations comprising:
receiving one or more input medical images acquired using one or more out-of-distribution image acquisition parameters and having out-of-distribution imaging properties, the one or more out-of-distribution image acquisition parameters and the out-of-distribution imaging properties being out-of-distribution with respect to training data on which the machine learning based model is trained;
generating one or more synthesized medical images from the one or more input medical images using a machine learning based generator network, the one or more synthesized medical images being generated for one or more in-distribution image acquisition parameters and having in-distribution imaging properties, the one or more in-distribution image acquisition parameters and the in-distribution imaging properties being in-distribution with respect to the training data on which the machine learning based model is trained, the machine learning based generator network comprising 1) an encoder network, conditioned based on values of at least one of the one or more out-of-distribution image acquisition parameters, for encoding the one or more input medical images into feature representations and 2) a decoder network, conditioned based on values of at least one of the one or more in-distribution image acquisition parameters, for decoding the feature representations to generate the one or more synthesized medical images;
performing the medical imaging analysis task based on the one or more synthesized medical images using the machine learning based model; and
outputting results of the medical imaging analysis task.
14 . The non-transitory computer readable medium of claim 13 , wherein the out-of-distribution image acquisition parameters and the in-distribution image acquisition parameters comprise parameters of an MRI (magnetic resonance imaging) scanner.
15 . The non-transitory computer readable medium of claim 13 , wherein the machine learning based generator network is jointly trained with another machine learning based generator network, the other machine learning based generator network generating synthesized out-of-distribution medical images from in-distribution training medical images.
16 . The non-transitory computer readable medium of claim 15 , wherein the machine learning based generator network generates reconstructed in-distribution medical images from the synthesized out-of-distribution medical images and wherein the machine learning based generator network is trained based on segmentation predictions from the reconstructed in-distribution medical images generated using the machine learning based model.
17 . The non-transitory computer readable medium of claim 13 , wherein the one or more input medical images comprises one or more mpMRI (multi-parametric magnetic resonance imaging) images.
18 . The non-transitory computer readable medium of claim 13 , wherein the one or more input medical images depict a prostate of a patient and the medical imaging analysis task is detection of prostate cancer.