| CPC A61B 5/4821 (2013.01) [A61B 5/374 (2021.01); A61B 5/7225 (2013.01); A61B 5/7264 (2013.01)] | 6 Claims |
|
1. A device for assessing anesthesia of an animal subject in real-time from electroencephalography (EEG) measurements of the animal subject during application of anesthesia, the device comprising:
a signal pre-processor comprising a first filter with a first filtering frequency band, a second filter with a second filtering frequency band, and a down-sampler to pre-process sequentially an incoming measured EEG signal from the EEG measurements of the animal subject;
wherein the pre-processing comprises:
removing unwanted signals and noise from the incoming measured EEG signal using the first filter and the second filter to generate a filtered incoming measured EEG signal; and
down-sampling the filtered incoming measured EEG signal by the down-sampler to generate a down-sampled filtered incoming measured EEG signal;
an epoch generator configured to generate an epoch signal containing epochs of 1-second epoch duration each and 10% overlapping of consecutive epochs;
wherein the epoch generator comprises a two-input multiplexer, an address generator, and a memory;
wherein the two-input multiplexer is configured to:
receive the down-sampled filtered incoming measured EEG signal, a feedback of the epoch signal, and a selector input from the address generator; and
generate a multiplexer output from the down-sampled filtered incoming measured EEG signal or the feedback of the epoch signal based on the selector input;
wherein the address generator comprises:
a slow counter configured to generate write addresses, and count from a slow counter start count to a slow counter end count in repeating slow count cycles; and
a fast counter configured to generate read addresses, and count from a fast counter start count to a fast counter end count in repeating fast count cycles that starts after each of the slow count cycles;
wherein the slow counter is further configured to repeat each of the slow count cycles after each of the fast count cycles;
wherein the fast counter is further configured to repeat each of the fast count cycles after each of the slow count cycles;
wherein the address generator is further configured to:
generate the selector input to the multiplexer to generate the multiplexer output from the down-sampled filtered incoming measured EEG signal during each of the slow count cycles and from the feedback of the epoch signal during each of the fast count cycles;
wherein the epoch generator is configured to:
receive the multiplexer output, the write addresses, and the read addresses;
store the received multiplexer output as data according to the received write addresses; and
retrieve data according to the received read addresses and output the retrieved data as a memory output;
wherein the epoch generator is configured to generate the epoch signal and output the epoch signal from the memory output;
a feature extractor configured to:
receive the epoch signal from the epoch generator;
extract a derivative feature of each epoch of the received epoch signal by computing a mean of accumulated squared-differences among epochs in the received epoch signal; and
extract a variance feature of each epoch of the received epoch signal by computing a square root of an average squared-deviation of epochs in the received epoch signal;
a classifier configured to:
perform a feature mapping on each derivative feature and each variance feature to its original value and its squared value; and
for each epoch of the received epoch signal, set a classification boundary from an output of the respective feature mappings to generate a classifier output for the epoch the classifier output being either that the animal subject is awake or that the animal subject is anesthetized; and
a predictor comprising a predictor circuit configured to:
accumulate a plurality of the classifier outputs corresponding to the epochs of the received epoch signal; and and
predict a likelihood of the animal subject being under anesthesia state from the accumulated classifier outputs.
|