Processing brain data using autoencoder neural networks

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing brain data using autoencoder neural networks. One of the methods includes obtaining brain data captured by one or more sensors characterizing brain activity of a patient; processing the brain data to generate modified brain data that characterizes a predicted local effect of a future treatment on the brain of the patient; processing the modified brain data using an autoencoder neural network to generate reconstructed brain data; and determining, using the reconstructed brain data, a predicted global effect of the future treatment on the brain of the patient.

What is claimed is: 
     
         1 - 20 . (canceled) 
     
     
         21 . A method comprising:
 obtaining desired brain data representing desired brain activity of a patient, wherein the desired brain data characterizes a global effect of a future treatment on the brain of the patient, wherein the global effect is an effect of the treatment on one or more regions of the brain that are not local to a target location of the future treatment in the brain; and   processing the desired brain data using an inverted autoencoder neural network to generate roadmap brain data that characterizes a predicted local effect of the future treatment on the brain of the patient, wherein the predicted local effect is a prediction of an effect of the future treatment on one or more regions of the brain that are local to the target location of the future treatment,   wherein:
 the inverted autoencoder neural network having been determined from a trained autoencoder neural network, the determining comprising inverting a plurality of operations of the autoencoder neural network; 
 the autoencoder neural network having been configured through training to process input brain data and to generate reconstructed brain data that is a prediction of the input brain data, the processing comprising:
 processing a network input generated from the input brain data using an encoder subnetwork to generate an embedding of the network input, and 
 processing the embedding of the network input using a decoder subnetwork to generate the reconstructed brain data; and 
 
 the autoencoder neural network having been trained by performing operations comprising, at each of a plurality of training time steps:
 obtaining second input brain data captured by one or more sensors characterizing brain activity of a second patient; 
 generating, from the second input brain data, a training network input; 
 processing, according to current values of a plurality of network parameters of the autoencoder neural network, the training network input to generate second reconstructed brain data, and 
 determining an update to the current values of the plurality of network parameters according to an error between i) the second input brain data and ii) the second reconstructed brain data. 
 
   
     
     
         22 . The method of  claim 21 , further comprising:
 determining, from the roadmap brain data, respective values for one or more parameters of the future treatment in order to actualize the desired brain activity of the patient; and   providing data characterizing the determined values for the one or more parameters to a healthcare system for providing the future treatment to the patient.   
     
     
         23 . The method of  claim 22 , wherein determining respective values for one or more parameters of the future treatment comprises:
 identifying one or more differences between (i) the roadmap brain data and (ii) real brain data captured by one or more second sensors characterizing real brain activity of the patient; and   determining the respective values for the one or more parameters of the future treatment according to the one or more identified differences.   
     
     
         24 . The method of  claim 22 , wherein the one or more parameters of the future treatment comprise one or more of:
 a recommended target location of the future treatment,   a recommended strength of the future treatment,   a recommended dose of the future treatment; or   a recommended schedule of the future treatment.   
     
     
         25 . The method of  claim 21 , wherein obtaining desired brain data characterizing desired brain activity of a patient comprises determining the desired brain data using one or more of:
 real brain data captured by one or more second sensors characterizing real brain activity of the patient; or   brain data of a third patient that characterizes the desired brain activity.   
     
     
         26 . The method of  claim 21 , wherein:
 the roadmap data characterizes a predicted local effect of the future treatment on the brain of the patient if the future treatment were provided according to first values for one or more parameters of the future treatment; and   the method further comprises:
 obtaining, from the inverted autoencoder neural network, one or more sets of second roadmap brain data that each characterize a predicted local effect of the future treatment on the brain of the patient if the future treatment were provided according to respective second values of the one or more parameters of the future treatment; and 
 selecting particular values for the one or more parameters of the future treatment using the roadmap brain data and the one or more sets of second roadmap data. 
   
     
     
         27 . The method of  claim 21 , wherein the future treatment is a transcranial magnetic stimulation (TMS) treatment. 
     
     
         28 . A system comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
 obtaining desired brain data representing desired brain activity of a patient, wherein the desired brain data characterizes a global effect of a future treatment on the brain of the patient, wherein the global effect is an effect of the treatment on one or more regions of the brain that are not local to a target location of the future treatment in the brain; and   processing the desired brain data using an inverted autoencoder neural network to generate roadmap brain data that characterizes a predicted local effect of the future treatment on the brain of the patient, wherein the predicted local effect is a prediction of an effect of the future treatment on one or more regions of the brain that are local to the target location of the future treatment,   wherein:
 the inverted autoencoder neural network having been determined from a trained autoencoder neural network, the determining comprising inverting a plurality of operations of the autoencoder neural network; 
 the autoencoder neural network having been configured through training to process input brain data and to generate reconstructed brain data that is a prediction of the input brain data, the processing comprising:
 processing a network input generated from the input brain data using an encoder subnetwork to generate an embedding of the network input, and 
 processing the embedding of the network input using a decoder subnetwork to generate the reconstructed brain data; and 
 
 the autoencoder neural network having been trained by performing operations comprising, at each of a plurality of training time steps:
 obtaining second input brain data captured by one or more sensors characterizing brain activity of a second patient; 
 generating, from the second input brain data, a training network input; 
 processing, according to current values of a plurality of network parameters of the autoencoder neural network, the training network input to generate second reconstructed brain data, and 
 determining an update to the current values of the plurality of network parameters according to an error between i) the second input brain data and ii) the second reconstructed brain data. 
 
   
     
     
         29 . The system of  claim 28 , the operations further comprising:
 determining, from the roadmap brain data, respective values for one or more parameters of the future treatment in order to actualize the desired brain activity of the patient; and   providing data characterizing the determined values for the one or more parameters to a healthcare system for providing the future treatment to the patient.   
     
     
         30 . The system of  claim 29 , wherein determining respective values for one or more parameters of the future treatment comprises:
 identifying one or more differences between (i) the roadmap brain data and (ii) real brain data captured by one or more second sensors characterizing real brain activity of the patient; and   determining the respective values for the one or more parameters of the future treatment according to the one or more identified differences.   
     
     
         31 . The system of  claim 29 , wherein the one or more parameters of the future treatment comprise one or more of:
 a recommended target location of the future treatment,   a recommended strength of the future treatment,   a recommended dose of the future treatment; or   a recommended schedule of the future treatment.   
     
     
         32 . The system of  claim 28 , wherein obtaining desired brain data characterizing desired brain activity of a patient comprises determining the desired brain data using one or more of:
 real brain data captured by one or more second sensors characterizing real brain activity of the patient; or   brain data of a third patient that characterizes the desired brain activity.   
     
     
         33 . The system of  claim 28 , wherein:
 the roadmap data characterizes a predicted local effect of the future treatment on the brain of the patient if the future treatment were provided according to first values for one or more parameters of the future treatment; and   the operations further comprise:
 obtaining, from the inverted autoencoder neural network, one or more sets of second roadmap brain data that each characterize a predicted local effect of the future treatment on the brain of the patient if the future treatment were provided according to respective second values of the one or more parameters of the future treatment; and 
 selecting particular values for the one or more parameters of the future treatment using the roadmap brain data and the one or more sets of second roadmap data. 
   
     
     
         34 . The system of  claim 28 , wherein the future treatment is a transcranial magnetic stimulation (TMS) treatment. 
     
     
         35 . One or more non-transitory computer storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:
 obtaining desired brain data representing desired brain activity of a patient, wherein the desired brain data characterizes a global effect of a future treatment on the brain of the patient, wherein the global effect is an effect of the treatment on one or more regions of the brain that are not local to a target location of the future treatment in the brain; and   processing the desired brain data using an inverted autoencoder neural network to generate roadmap brain data that characterizes a predicted local effect of the future treatment on the brain of the patient, wherein the predicted local effect is a prediction of an effect of the future treatment on one or more regions of the brain that are local to the target location of the future treatment,   wherein:
 the inverted autoencoder neural network having been determined from a trained autoencoder neural network, the determining comprising inverting a plurality of operations of the autoencoder neural network; 
 the autoencoder neural network having been configured through training to process input brain data and to generate reconstructed brain data that is a prediction of the input brain data, the processing comprising:
 processing a network input generated from the input brain data using an encoder subnetwork to generate an embedding of the network input, and 
 processing the embedding of the network input using a decoder subnetwork to generate the reconstructed brain data; and 
 
 the autoencoder neural network having been trained by performing operations comprising, at each of a plurality of training time steps:
 obtaining second input brain data captured by one or more sensors characterizing brain activity of a second patient; 
 generating, from the second input brain data, a training network input; 
 processing, according to current values of a plurality of network parameters of the autoencoder neural network, the training network input to generate second reconstructed brain data, and 
 determining an update to the current values of the plurality of network parameters according to an error between i) the second input brain data and ii) the second reconstructed brain data. 
 
   
     
     
         36 . The non-transitory computer storage media of  claim 35 , the operations further comprising:
 determining, from the roadmap brain data, respective values for one or more parameters of the future treatment in order to actualize the desired brain activity of the patient; and   providing data characterizing the determined values for the one or more parameters to a healthcare system for providing the future treatment to the patient.   
     
     
         37 . The non-transitory computer storage media of  claim 36 , wherein determining respective values for one or more parameters of the future treatment comprises:
 identifying one or more differences between (i) the roadmap brain data and (ii) real brain data captured by one or more second sensors characterizing real brain activity of the patient; and   determining the respective values for the one or more parameters of the future treatment according to the one or more identified differences.   
     
     
         38 . The non-transitory computer storage media of  claim 36 , wherein the one or more parameters of the future treatment comprise one or more of:
 a recommended target location of the future treatment,   a recommended strength of the future treatment,   a recommended dose of the future treatment; or   a recommended schedule of the future treatment.   
     
     
         39 . The non-transitory computer storage media of  claim 35 , wherein obtaining desired brain data characterizing desired brain activity of a patient comprises determining the desired brain data using one or more of:
 real brain data captured by one or more second sensors characterizing real brain activity of the patient; or   brain data of a third patient that characterizes the desired brain activity.   
     
     
         40 . The non-transitory computer storage media of  claim 35 , wherein:
 the roadmap data characterizes a predicted local effect of the future treatment on the brain of the patient if the future treatment were provided according to first values for one or more parameters of the future treatment; and   the operations further comprise:
 obtaining, from the inverted autoencoder neural network, one or more sets of second roadmap brain data that each characterize a predicted local effect of the future treatment on the brain of the patient if the future treatment were provided according to respective second values of the one or more parameters of the future treatment; and 
 selecting particular values for the one or more parameters of the future treatment using the roadmap brain data and the one or more sets of second roadmap data. 
   
     
     
         41 . The non-transitory computer storage media of  claim 35 , wherein the future treatment is a transcranial magnetic stimulation (TMS) treatment.