Case-based reasoning in the cloud using deep learning

A method for identifying first data stored at a first computing device based on second data generated by and received from a second computing device, the first data being related to a plurality of first patients and the second data being related to a second patient, the method comprising:

• receiving, by a processor of the first computing device, the second data from the second computing device, wherein the second data comprises a first data signature, the first data signature being related to the second patient and representing a plurality of modality-specific datasets, wherein the first data signature has a first dimensionality and a modality-specific dataset of the plurality of modality-specific datasets has a second dimensionality, the first dimensionality being lower than the second dimensionality;
• comparing, by the processor, the first data signature with a plurality of second data signatures stored at the first computing device, wherein the plurality of second data signatures is related to the plurality of first patients, respectively, wherein the first data signature includes two values representing the second patient, and wherein each second data signature of the plurality of second data signatures includes two values representing a patient of the plurality of first patients, respectively, and wherein the comparing comprises determining a similarity representation between the first data signature and each second data signature of the plurality of second data signatures;
• identifying, by the processor, the first data based on the comparison, wherein the first data comprises at least one second data signature of the plurality of second data signatures, the at least one second data signature corresponding to the at least one first patient of the plurality of first patients, respectively, and wherein the identifying further comprises identifying the at least one second data signature based on a comparison of the determined similarity representations to a threshold;
• and
• sending, by the processor, at least a portion of the first data to the second computing device.

A computing device for presenting data related to a patient, the computing device comprising:

• a processor configured to:
• identify a plurality of modality-specific datasets, the plurality of modality-specific datasets being related to the patient, a first modality-specific dataset of the plurality of modality-specific datasets having a first dimensionality and a second modality-specific dataset of the plurality of modality-specific datasets having a second dimensionality, and the first dimensionality being different than the second dimensionality;
• train, by the processor, an encoder, wherein the encoder is a neural network, wherein the neural network comprises modality-specific sub-networks and a joint sub-network, outputs of the modality-specific sub-networks being inputs to the joint sub-network, wherein at least one of the modality-specific sub-networks connects with the other modality-specific sub-networks at a different level within the joint sub-network, wherein the training results in a trained encoder, and wherein training the encoder comprises:
• identification, by the processor, of input data and a target value associated with the input data;
• and
• optimization, by the processor, of the neural network based on the identified input data and target value;
• generate, with the trained encoder, a first data signature for the first modality-specific dataset and a second data signature for the second modality-specific dataset, a dimensionality of the first data signature being lower than the first dimensionality and a dimensionality of the second data signature being lower than the second dimensionality, the dimensionality of the first data signature being the same as the dimensionality of the second data signature;
• generate, with the trained encoder, a joint data signature based on the first data signature and the second data signature as inputs to the trained encoder;
• transmit the joint data signature to another computing device, the other computing device being in communication with the computing device via a network;
• and
• receive the data related to the patient from the other computing device, the data related to the patient also being related to the transmitted joint data signature;
• and
• a display in communication with the processor, the display being configured to display a representation of the received data.

A computing device for identifying data related to a first patient, the computing device comprising:

• a memory configured to store a plurality of first data signatures, each first data signature of the plurality of first data signatures being related to a respective second patient;
• a processor in communication with the memory, the processor being configured to:
• receive a second data signature from another computing device, the second data signature being related to the first patient, the other computing device being in communication with the computing device via a network, the second data signature representing a plurality of modality-specific datasets, the second data signature having a first dimensionality, and a modality-specific dataset of the plurality of modality-specific datasets having a second dimensionality, the first dimensionality being lower than the second dimensionality, the second data signature including two values representing the first patient, and each first data signature of the plurality of first data signatures including two values representing a patient of the plurality of second patients, respectively;
• compare the second data signature with the plurality of first data signatures stored at the memory;
• determine a similarity representation between the second data signature and each first data signature of the plurality of first data signatures;
• identify the data related to the first patient from the plurality of first data signatures based on the comparison, the data related to the first patient comprising at least one first data signature of the plurality of first data signatures;
• identify the at least one first data signature based on a comparison of the determined similarity representations to a threshold;
• and
• send at least a portion of the data related to the first patient to the other computing device.