| CPC A61B 5/7282 (2013.01) [A61B 5/384 (2021.01); A61B 5/7267 (2013.01); G06N 20/00 (2019.01); G16H 50/20 (2018.01)] | 20 Claims |
|
1. A stroke detection device comprising:
a data store comprising a program of instructions, wherein the data store comprises:
a feature extraction engine configured to extract features from a rolling window of electroencephalogram (EEG) data to generate a 1-D input vector (135) of the extracted features; and,
a classification engine configured to apply the 1-D input vector to an ensemble stroke classification model (ESCM); and,
a processor operably coupled to the data store such that, when the processor executes the program of instructions, the processor causes operations to be performed to automatically and accurately predict whether a patient is experiencing a stroke, the operations comprising:
receive a finite time window of a first predetermined duration of EEG data from a monitoring device operably coupled to the patient;
extract, by the feature extraction engine, at least four classes of features from the received EEG data;
aggregate the extracted features into the 1-D input vector;
apply the classification engine to the 1-D input vector such that the ESCM operates on the extracted features aggregated into the 1-D input vector;
generate, by the ESCM, a binary stroke prediction; and,
generate and transmit a prediction signal to a user interface device such that an indication of the binary stroke prediction is provided to a user,
wherein:
the ESCM comprises:
a class-specific model set for each of the classes of features, each set comprising multiple class-specific models for each of a corresponding group of architectures, each class-specific model configured to receive the 1-D input vector and operate on features of the corresponding class; and,
a general model set including multiple general models for each of a corresponding group of architectures, each general model configured to receive the 1-D input vector and operate on features of multiple of the classes; and,
the binary stroke prediction is generated based on a predetermined weighted aggregation of an output of each of the class-specific models and each of the general models, such that a stroke prediction with an area under a receiver operating characteristic curve greater than 0.95 is determined in a finite time window of less than 10 minutes.
|