Medical event prediction using a personalized dual-channel combiner network

Abstract

Systems and methods for predicting an occurrence of a medical event for a patient using a trained neural network. Historical patient data is preprocessed to generate normalized training samples, and the normalized training samples are sent to a personalized deep convolutional neural network for model pretraining and updating of model parameters. The pretrained model is stored in a remote server for utilization by a local machine for personalization during a preparation time period for a medical treatment. A normalized finetuning set is generated as output, and the model parameters are iteratively finetuned. A personal prediction score for future medical events is generated, and an operation of a medical treatment device is controlled responsive to the prediction score.

What is claimed is: 
     
         1 . A computer implemented method for predicting an occurrence of a medical event for a patient using a trained neural network, comprising:
 preprocessing received historical patient data for a plurality of patients to generate a plurality of normalized training samples;   sending the normalized training samples to a personalized deep convolutional neural network (P-DCCN) and initiating model pretraining and updating of model parameters using the P-DCCN using the normalized training samples;   storing the pretrained model in a remote server for utilization for personalization by a local machine during a preparation time period for a medical treatment;   generating a normalized finetuning set as output from the P-DCCN by processing input personal data for the patient from the local machine;   iteratively finetuning the model parameters of the P-DCCN by performing a plurality of training iterations using the generated normalized finetuning set;   computing a probability of an incidence of a medical event by concatenating static and temporal representations, received from a static channel and a temporal channel, respectively, and computing the probability using a multilayer perceptron of the P-DCCN.   
     
     
         2 . The method of  claim 1 , wherein the P-DCCN is finetuned by optimizing using a regression loss function as follows: 
       
         
           
             
               l 
               = 
               
                 
                   
                     1 
                     N 
                   
                   ⁢ 
                   
                     
                       ∑ 
                         
                     
                     
                       i 
                       = 
                       1 
                     
                     N 
                   
                   ⁢ 
                   
                     
                        
                       
                         
                           
                             y 
                             ^ 
                           
                           i 
                         
                         - 
                         
                           y 
                           i 
                         
                       
                        
                     
                     2 
                     2 
                   
                 
                 + 
                 
                   λ 
                   ⁢ 
                   
                     
                        
                       θ 
                        
                     
                     2 
                     2 
                   
                 
               
             
           
         
         where y i  is a true indicator of an incidence of an event for an i-th sample in the training samples, ŷ i  is a predicted score for the i-th sample, N is a total number of the training samples, θ represents the model parameters, λ is a hyperparameter which controls a regularization on the model parameters to avoid overfitting during the training. 
       
     
     
         3 . The method of  claim 1 , wherein the preprocessing received historical patient data for a plurality of patients further comprises segmenting recording data and generating time series samples. 
     
     
         4 . The method of  claim 1 , wherein the P-DCCN includes a static channel, which is utilized for processing static and comparatively low frequency temporal features, and a temporal channel, which is utilized for processing comparatively high frequency temporal features. 
     
     
         5 . The method of  claim 4 , wherein the static channel includes a multilayer perceptron (MLP) configured to encode information in static features x s  to a compact representation of the static features h s  by:
     h   s   =f   MLP ( x   s )   where f MLP (⋅) represents multiple layers of a fully connected network with the form W s x s +b s , with W s  and b s  being model parameters to be trained.   
     
     
         6 . The method of  claim 4 , wherein the temporal channel includes a plurality of long short term memory (LSTM) layers for processing the temporal features, represented by a sequence of vectors x 1 , . . . , x T , to output a sequence of compact representations h 1 , . . . , h T  by:
     h   1   , . . . ,h   T   =f   LSTM ( x   1   , . . . ,x   T )   where f LSTM (⋅) includes multiple layers of LSTM units, which include trainable model parameters.   
     
     
         7 . A system for predicting an occurrence of a medical event for a patient using a trained neural network, comprising:
 a processor operatively coupled to a computer-readable storage medium, the processor being configured for:   preprocessing received historical patient data for a plurality of patients to generate a plurality of normalized training samples;   sending the normalized training samples to a personalized deep convolutional neural network (P-DCCN) and initiating model pretraining and updating of model parameters using the P-DCCN using the normalized training samples;   storing the pretrained model in a remote server for utilization for personalization by a local machine during a preparation time period for a medical treatment;   generating a normalized finetuning set as output from the P-DCCN by processing input personal data for the patient from the local machine;   iteratively finetuning the model parameters of the P-DCCN by performing a plurality of training iterations using the generated normalized finetuning set;   computing a probability of an incidence of a medical event by concatenating static and temporal representations, received from a static channel and a temporal channel, respectively, and computing the probability using a multilayer perceptron of the P-DCCN.   
     
     
         8 . The system of  claim 7 , wherein the P-DCCN is finetuned by optimizing using a regression loss function as follows: 
       
         
           
             
               l 
               = 
               
                 
                   
                     1 
                     N 
                   
                   ⁢ 
                   
                     
                       ∑ 
                         
                     
                     
                       i 
                       = 
                       1 
                     
                     N 
                   
                   ⁢ 
                   
                     
                        
                       
                         
                           
                             y 
                             ^ 
                           
                           i 
                         
                         - 
                         
                           y 
                           i 
                         
                       
                        
                     
                     2 
                     2 
                   
                 
                 + 
                 
                   λ 
                   ⁢ 
                   
                     
                        
                       θ 
                        
                     
                     2 
                     2 
                   
                 
               
             
           
         
         where y i  is a true indicator of an incidence of an event for an i-th sample in the training samples, ŷ i  is a predicted score for the i-th sample, N is a total number of the training samples, θ represents the model parameters, λ is a hyperparameter which controls a regularization on the model parameters to avoid overfitting during the training. 
       
     
     
         9 . The system of  claim 7 , wherein the preprocessing received historical patient data for a plurality of patients further comprises segmenting recording data and generating time series samples. 
     
     
         10 . The system of  claim 7 , wherein the P-DCCN includes a static channel, which is utilized for processing static and comparatively low frequency temporal features, and a temporal channel, which is utilized for processing comparatively high frequency temporal features. 
     
     
         11 . The system of  claim 10 , wherein the static channel includes a multilayer perceptron (MLP) configured to encode information in static features x s  to a compact representation of the static features h s  by:
     h   s   =f   MLP ( x   s )   where f MLP (⋅) represents multiple layers of a fully connected network with the form W s x s +b s , with W s  and b s  being model parameters to be trained.   
     
     
         12 . The system of  claim 10 , wherein the temporal channel includes a plurality of long short term memory (LSTM) layers for processing the temporal features, represented by a sequence of vectors x 1 , . . . , x T , to output a sequence of compact representations h 1 , . . . , h T  by:
     h   1   , . . . ,h   T   =f   LSTM ( x   1   , . . . ,x   T )   where f LSTM (⋅) includes multiple layers of LSTM units, which include trainable model parameters.   
     
     
         13 . A non-transitory computer-readable storage medium comprising a computer-readable program for predicting an occurrence of a medical event for a patient using a trained neural network, wherein the computer-readable program when executed on a computer causes the computer to perform the steps of:
 preprocessing received historical patient data for a plurality of patients to generate a plurality of normalized training samples;   sending the normalized training samples to a personalized deep convolutional neural network (P-DCCN) and initiating model pretraining and updating of model parameters using the P-DCCN using the normalized training samples;   storing the pretrained model in a remote server for utilization for personalization by a local machine during a preparation time period for a medical treatment;   generating a normalized finetuning set as output from the P-DCCN by processing input personal data for the patient from the local machine;   iteratively finetuning the model parameters of the P-DCCN by performing a plurality of training iterations using the generated normalized finetuning set;   computing a probability of an incidence of a medical event by concatenating static and temporal representations, received from a static channel and a temporal channel, respectively, and computing the probability using a multilayer perceptron of the P-DCCN.   
     
     
         14 . The computer-readable storage medium of  claim 13 , wherein the P-DCCN is finetuned by optimizing using a regression loss function as follows: 
       
         
           
             
               l 
               = 
               
                 
                   
                     1 
                     N 
                   
                   ⁢ 
                   
                     
                       ∑ 
                         
                     
                     
                       i 
                       = 
                       1 
                     
                     N 
                   
                   ⁢ 
                   
                     
                        
                       
                         
                           
                             y 
                             ^ 
                           
                           i 
                         
                         - 
                         
                           y 
                           i 
                         
                       
                        
                     
                     2 
                     2 
                   
                 
                 + 
                 
                   λ 
                   ⁢ 
                   
                     
                        
                       θ 
                        
                     
                     2 
                     2 
                   
                 
               
             
           
         
         where y i  is a true indicator of an incidence of an event for an i-th sample in the training samples, ŷ i  is a predicted score for the i-th sample, N is a total number of the training samples, θ represents the model parameters, λ is a hyperparameter which controls a regularization on the model parameters to avoid overfitting during the training. 
       
     
     
         15 . The computer-readable storage medium of  claim 13 , wherein the P-DCCN includes a static channel, which is utilized for processing static and comparatively low frequency temporal features, and a temporal channel, which is utilized for processing comparatively high frequency temporal features. 
     
     
         16 . The computer-readable storage medium of  claim 13 , wherein the temporal channel includes a plurality of long short term memory (LSTM) layers for processing the temporal features, represented by a sequence of vectors x 1 , . . . , x T , to output a sequence of compact representations h 1 , . . . , h T  by:
     h   1   , . . . ,h   T   =f   LSTM ( x   1   , . . . ,x   T )   where f LSTM (⋅) includes multiple layers of LSTM units, which include trainable model parameters.   
     
     
         17 . The computer-readable storage medium of  claim 16 , further comprising computing a probability of an incidence of a medical event by concatenating static and temporal representations h s  and h d , received from a static channel and a temporal channel, respectively, and computing the probability using a multilayer perceptron (MLP) by:
     ŷ=f   MLP ([ h   s   ,h   d ])   where ŷ is a score which indicates the probability of the incidence of the medical event.