Loading patent…
AssigneeShanghai United Imaging Intelligence Co., Ltd.
InventorsAbhishek Sharma, Arun Innanje, Ziyan Wu, Shanhui Sun, Terrence Chen
Estimating the term…
Methods and systems for enhancing a distributed medical network. For example, a computer-implemented method includes inputting training data corresponding to each local computer into their corresponding machine learning model; generating a plurality of local losses including generating a local loss for each machine learning model based at least in part on the corresponding training data; generating a plurality of local parameter gradients including generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss; generating a global parameter update based at least in part on the plurality of local parameter gradients; and updating each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update.
A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:
A system for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the system comprising:
A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:
A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:
A non-transitory computer-readable medium with instructions stored thereon, that when executed by a processor, causes the processor to perform one or more processes including:
A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:
Dependencies are inferred from the claim text.
Drawings are temporarily unavailable. Open the PDF at USPTO
From the US filing history. Foreign family members aren’t listed.
Learning methods: backpropagation, e.g. using gradient descent
In which an application is distributed across nodes in the network: for distributed storage of data in networks, e.g. transport arrangements for…
Architecture, e.g. interconnection topology: combinations of networks
Architecture, e.g. interconnection topology: convolutional networks [CNN, ConvNet]
About 13 years 10 months of term left · ends 27 Aug 2040 (estimated)
Hospitals collectively have a great amount of patient medical data that can be used for various tasks. For example, each hospital may include a medical system configured for a certain medical application, which may utilize particular patient medical data to improve their ability to perform a particular task. Currently, hospitals, which may be located at various geographical locations, suffer from great inefficiencies in sharing pertinent patient medical data with each other, in particularly data for improving neural network. Thus, systems and methods for enhancing data sharing in medical data systems are desirable.
Certain embodiments of the present invention are directed to medical data systems. More particularly, some embodiments of the invention provide methods and systems for enhancing a distributed medical network. Merely by way of example, some embodiments of the invention are configured to train a plurality of machine learning models hosted by a plurality of local hospitals. But it would be recognized that the invention has a much broader range of applicability.
In various embodiments, a computer-implemented method for enhancing a distributed medical network (e.g., a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set) includes: inputting training data corresponding to each local computer into their corresponding machine learning model; generating a plurality of local losses including generating a local loss for each machine learning model based at least in part on the corresponding training data; generating a plurality of local parameter gradients including generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss; generating a global parameter update based at least in part on the plurality of local parameter gradients; and updating each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update.
FIG. 1 is a simplified diagram showing a system for enhancing a distributed medical network, according to some embodiments.
FIGS. 2A and 2B are simplified diagrams showing a method for enhancing a centralized distributed medical network, according to some embodiments.
FIGS. 3A and 3B are simplified diagrams showing a method for enhancing a decentralized distributed medical network, according to some embodiments.
FIG. 1 is a simplified diagram showing a system for enhancing a distributed medical network, according to some embodiments. This diagram is merely an example, which should not unduly limit the scope of the claims. One of ordinary skill in the art would recognize many variations, alternatives, and modifications. In some examples, the system 10 includes a connection establishing module 12, a data inputting module 14, a loss generating module 16, a gradient generating module 18, a parameter set generating module 20, a gradient or parameter delivering module 22, an update generating module 24, and a model updating module 26. In certain examples, the system 10 further includes an update delivering module 28, a time out module 30, and/or a compression and encryption module 32. In some examples, the system 10 is configured to enhance a distributed medical network. For example, the system 10 is configured to enhance a distributed medical network including a plurality of local computers, where each local computer hosts a machine learning model configured with an active parameter set. In some examples, the distributed medical network further includes a central computer, (e.g., coupled to the plurality of local computers). In certain examples, the system 10 is configured to implement method S100 of FIGS. 2A and 2B and/or method S200 of FIGS. 3A and 3B. Although the above has been shown using a selected group of components, there can be many alternatives, modifications, and variations. For example, some of the components may be expanded and/or combined. Some components may be removed. Other components may be inserted to those noted above. Depending upon the embodiment, the arrangement of components may be interchanged with others replaced.
In various embodiments, the connection establishing module 12 is configured to establish connection between the plurality of local computers and a central computer. In some examples, the connection establishing module 12 is configured to establish connection between each local computer of the plurality of local computers. In various examples, the connection establishing module 12 is configured to establish connection by a wireless and/or wired connection. In certain examples, the connection establishing module 12 is configured to establish a data transfer connection for sharing, transferring, transmitting, and/or communicating data. In certain examples, the connection establishing module 12 is a separate module coupled to the system 10 and not part of system 10.
In various embodiments, the data inputting module 14 is configured to input data into each local computer of the plurality of local computers. In some examples, the data inputting module 14 is configured to input priming data and/or training data into each machine learning model hosted at leach local computer of the plurality of local computers. For example, the data inputting module 14 is configured to input priming data and/or training data corresponding to each local computer into their corresponding machine learning model. In certain examples, the data inputting module 14 is configured to input a first training data into a first machine learning model hosted at a first local computer of the plurality of local computers, a second training data into a second machine learning model hosted at a second local computer of the plurality of local computers, . . . , and a Nth training data into a Nth machine learning model hosted at a Nth local computer of the plurality of local computers. In certain examples, training data includes a medical image (e.g., a pre-analyzed medical image with a corresponding ground truth) and/or patient data. In some examples, the training data inputted into each machine learning model is different. In certain examples, the data inputting module 14 is configured to input a first priming data into a first machine learning model hosted at a first local computer of the plurality of local computers, a second priming data into a second machine learning model hosted at a second local computer of the plurality of local computers, . . . , and a Nth priming data into a Nth machine learning model hosted at a Nth local computer of the plurality of local computers. In certain examples, priming data includes a medical image (e.g., a pre-analyzed medical image with a corresponding ground truth) and/or patient data. In some examples, the priming data inputted into each machine learning model is different. In various examples, the training data and priming data for each machine learning model is the same.