| CPC G16H 70/20 (2018.01) [G16H 70/60 (2018.01)] | 9 Claims |
|
1. A medical information processing apparatus comprising:
a storage device configured to store a medical knowledge graph including nodes corresponding to medical care events, and edges indicative of a relationship between the nodes, a graph feature of the medical knowledge graph being expressed by a mathematical model characterized by patient background information, the graph feature including a kind of a node, a value of a node, a kind of an edge, and/or a value of an edge; and
processing circuitry configured to obtain patient background information relating to a background factor of one or more target patients, configured to compute variation of the graph feature relating to the target patient, based on the patient background information of the target patient and the mathematical model, and configured to display the variation of the graph feature on a display device,
wherein the variation of the graph feature is a correlation degree of the graph feature with the patient background information, and
the processing circuitry is configured to compute the variation of the graph feature in such a manner that:
if the one or more target patients comprise a plurality of target patients, the processing circuitry computes the graph feature for each of the target patients based on the patient background information and the mathematical model and computes, as the variation of the graph feature, an index of dispersion of the graph features over the target patients, and
if the one or more target patients comprise a single target patient, the processing circuitry computes the graph feature for the patient background information at each of a plurality of time points based on the patient background information and the mathematical model and computes, as the variation of the graph feature, an index of dispersion of the graph features over the time points,
wherein the processing circuitry is configured to determine parameters of the mathematical model by inputting a patient graph, which includes a variation, to a machine learning model and output a determination result, the machine learning model having been trained using patient graphs and determination result labels in association with each other,
wherein the machine learning model is a neural network that includes at least
a convolution layer configured to receive the inputted patient graph, executes a graph convolution process that applies learned weight parameters to nodes of the patient graph, and outputs a post-convolution patient graph,
a readout layer configured to convert the post-convolution patient graph to a feature vector having a same number of dimensions as a number of nodes in the post-convolution patient graph, and
a dense layer configured to convert the feature vector into disease classification information as the determination result, the outputted determination result being a numerical value indicative of a probability of relevance to each of one or more relevant diseases.
|