US5598508A — Real-time waveform analysis using artificial neur…

Claim 1

Independent Claim 201 words

A method of analyzing a periodic characteristic waveform in a signal using neural networks, said method comprising the steps of: (a) protraining first and second neural networks to recognize a starting point and an end point of said characteristic waveform; (b) transmitting said signal to a buffer, said buffer having greater than N storage locations each having an index number, said characteristic waveform in said signal having starting and end points: (c) serially transmitting the contents of said buffer to said first and second neural networks each neural network having N input nodes, and each input node having an index number, inputting data to each input node from buffer locations having the same index number as a corresponding input node, said neural network input nodes receiving a value of zero from buffer locations not yet receiving any portions of said signal; (d) causing said first neural network to generate an output which is a representation of the index number of the input node receiving said starting point of said characteristic waveform, and causing said second neural network to generate an output which is a representation of the index number of the input node receiving said end point of said characteristic waveform.

Claim 2

Independent Claim 201 words

A method of analyzing a periodic characteristic waveform in a signal using neural networks, said method comprising the steps of: (a) protraining first and second neural networks to recognize a starting point and an end point of said characteristic waveform; (b) transmitting said signal to a buffer, said buffer having greater than N storage locations each having an index number, said characteristic waveform in said signal having starting and end points: (c) serially transmitting the contents of said buffer to said first and second neural networks each neural network having N input nodes, and each input node having an index number, inputting data to each input node from buffer locations having the same index number as a corresponding input node, said neural network input nodes receiving a value of zero from buffer locations not yet receiving any portions of said signal; (d) causing said first neural network to generate an output which is a representation of the index number of the input node receiving said starting point of said characteristic waveform, and causing said second neural network to generate an output which is a representation of the index number of the input node receiving said end point of said characteristic waveform.

Claim 3

Dependent Claim 24 words

The method of claim 1 further comprising the steps of assigning a negative value to buffer not yet receiving a portion of said signal.

Claim 4

Dependent Claim 79 words

The method of claim 1 wherein step (d) includes the steps of: determining if said end point chosen is a false end point by determining if the index number of said end point is one of the buffer locations which has not yet received any portion of said signal and if so, repeating steps (a)-(d) until an end point is produced by said second neural network which is an index number that has received a portion of said signal.

Claim 5

Dependent Claim 64 words

The method of claim 3 further comprising the step of determining if the end point is a predetermined distance from the N th index and if so, shifting the waveform toward lower index locations in said buffer, whereby more of a terminal portion of the waveform may be analyzed because it does not fall outside a buffer location corresponding to the N th index.

Claim 6

Dependent Claim 36 words

The method of claim 1 further comprising the steps of: clearing the buffer of all data between said starting point and end point, and repeating steps (a)-(d) to analyze the next consecutive waveform in said signal.

Claim 7

Independent Claim 185 words

A method of analyzing a periodic characteristic waveform in a signal using a neural network, said characteristic waveform belonging to one of a plurality of predetermined classes, said method comprising the steps of: (a) pretraining a neural network to recognize a starting point and an end point of said characteristic waveform; (b) transmitting said signal to a buffer, said buffer having greater than N storage locations each having an index number, said characteristic waveform in said signal having a starting point and an end point; (c) serially transmitting the contents of said buffer to said neural network, the neural network having N input nodes, each input node having an index number, inputting data to each input node from buffer locations having the same index number as a corresponding input node, said neural network input nodes receiving a value of zero from buffer locations not yet receiving any portion of said signal; (d) generating first and second outputs by said neural network which are a representation of the index number of the input node receiving the starting point and end point respectively of said characteristic waveform.

Claim 8

Independent Claim 185 words

A method of analyzing a periodic characteristic waveform in a signal using a neural network, said characteristic waveform belonging to one of a plurality of predetermined classes, said method comprising the steps of: (a) pretraining a neural network to recognize a starting point and an end point of said characteristic waveform; (b) transmitting said signal to a buffer, said buffer having greater than N storage locations each having an index number, said characteristic waveform in said signal having a starting point and an end point; (c) serially transmitting the contents of said buffer to said neural network, the neural network having N input nodes, each input node having an index number, inputting data to each input node from buffer locations having the same index number as a corresponding input node, said neural network input nodes receiving a value of zero from buffer locations not yet receiving any portion of said signal; (d) generating first and second outputs by said neural network which are a representation of the index number of the input node receiving the starting point and end point respectively of said characteristic waveform.

Claim 9

Dependent Claim 109 words

The method of claim 6 further comprising the step of pretraining a second neural network having N input nodes and R output nodes, to produce signals in said output nodes which represent a plurality of classifications of a plurality of characteristic waveforms in said signal, and one of said classifications indicating a partial wave, signifying that a complete characteristic wave form has not been received by said second neural network; and transmitting said characteristic waveform to said N input nodes of said second neural network, wherein said second neural network produces an output which classifies said waveform as either a partial waveform or as one of said characteristic waveforms.

Claim 10

Dependent Claim 33 words

The method of claim 7 further comprising the step of: normalizing said waveform to comprise a predetermined duration and amplitude, wherein said starting point and end point of said waveform comprise N samples.

Claim 11

Dependent Claim 28 words

The method of claim 8 further comprising the step of normalizing lead and lag portions of said waveform preceding said starting point and following said end point respectively.

Claim 12

Dependent Claim 156 words

A method of claim 6 further comprising the steps of: using a classification means to classify said waveform based on its shape, said classification assigning numerical values to a likelihood that said waveform belongs to one of said plurality of predetermined classes; when said numerical values do not unambiguously indicate a single class, transmitting said numerical values of said classifications to input nodes of a second neural network; transmitting additional information relating to said waveform to additional input nodes of said second neural network: and training said second neural network to classify, said waveform by producing output numerical values representing a likelihood that said waveform belongs to one of a plurality of classes, training said second neural network with inputs containing said additional informations of said additional information serving to resolve any ambiguities between various classifications of said shape by correlating a typical shape of a particular waveform classification with particular values for said additional information.

Claim 13

Dependent Claim 48 words

The method of claim 10 wherein said waveform is a signal derived from a capnogram and said classes produced by said classification system represent abnormalities in said waveform and said additional information is transmitted to said neural network includes minimum CO 2 maximum CO 2 and respiratory rate.

Claim 14

Independent Claim 226 words

A system for analyzing a periodic characteristic waveform in a signal, using neural networks, said characteristic waveform belonging to one of a plurality of predetermined classes, said system comprising: means for transmitting said signal to a buffer, said buffer having greater than N storage locations each having an index number, said characteristic waveform in said signal having a starting point and an end point; first and second neural networks each having N input nodes, and each input node having an index number, said first and second neural networks comprising pretrained neural networks which have been pretrained to recognize the starting and end points of said characteristic waveform; means for serially transmitting the contents of said buffer to said first and second neural networks, inputting data to each input mode from buffer locations having the same index number as a corresponding input node, said neural network input nodes receiving a value of zero from buffer location not yet receiving any portion of said signal; and means for causing said first neural network to generate an output which is a representation of the index number of the input node receiving the starting point of said characteristic waveform, and for causing said second neural network to generate and output which is a representation of the index number of the input node receiving said end point of said characteristic waveform.

Claim 15

Independent Claim 226 words

A system for analyzing a periodic characteristic waveform in a signal, using neural networks, said characteristic waveform belonging to one of a plurality of predetermined classes, said system comprising: means for transmitting said signal to a buffer, said buffer having greater than N storage locations each having an index number, said characteristic waveform in said signal having a starting point and an end point; first and second neural networks each having N input nodes, and each input node having an index number, said first and second neural networks comprising pretrained neural networks which have been pretrained to recognize the starting and end points of said characteristic waveform; means for serially transmitting the contents of said buffer to said first and second neural networks, inputting data to each input mode from buffer locations having the same index number as a corresponding input node, said neural network input nodes receiving a value of zero from buffer location not yet receiving any portion of said signal; and means for causing said first neural network to generate an output which is a representation of the index number of the input node receiving the starting point of said characteristic waveform, and for causing said second neural network to generate and output which is a representation of the index number of the input node receiving said end point of said characteristic waveform.

Claim 16

Dependent Claim 44 words

The system of claim 12 further comprising: means for determining if said end point chosen is a false end point because the index number of said end point is one of the buffer locations which had not yet received any portions of said signal.

Claim 17

Dependent Claim 117 words

The system of claim 12 further comprising; a third neural network having N input nodes and R output nodes, said third neural network having been trained to produce neural network signals in said output nodes which represent a plurality of classifications of said characteristic waveforms, and one of said classifications indicating a partial wave, signifying the a complete characteristic waveform has not been received by said neural network; and means for transmitting each of said neural network signals to one of said N input nodes of said pretrained third neural network, wherein said pretrained third neural network produces an out put which classifies said waveform as either a partial waveform or as one of said characteristic waveforms.

Claim 18

Dependent Claim 32 words

The system of claim 14, further comprising: means for normalizing said waveform to comprise a predetermined duration and amplitude, wherein said starting point and end point of said waveform comprise N samples.

Claim 19

Dependent Claim 27 words

The system of claim 15 further comprising means for normalizing lead and lag portions of said waveform preceding said starting point and following said end point respectively.

Claim 20

Dependent Claim 158 words

The system of claim 12 further comprising: a classification means for classifying said waveform based on its shape, said classification assigning numerical values to a likelihood that said waveform belongs to one of a plurality of predetermined classes; means for transmitting said numerical values of said classifications to input nodes of a third neural network when said numerical values do not unambiguously indicate a single class; means for transmitting additional information relating to said waveform to additional input nodes of said third neural network; and means for training said third neural network to classify said waveform by producing output numerical values representing a likelihood that said waveform belongs to one of a plurality of classes, said third neural network being trained with inputs containing said additional information, said additional information serving to resolve any ambiguities between various classifications of said shape, by correlating a typical shape of a particular waveform classification with particular values for said additional information.

Claim 21

Dependent Claim 46 words

The system of claim 17 wherein said waveform is a signal derived from a capnogram, said classes produced by said classification system represent abnormalities in said waveform, said additional information is transmitted to said neural network includes minimum CO 2 maximum CO 2 and respiratory rate.

Claim 22

Independent Claim 128 words

A method of analyzing a periodic characteristic waveform in a signal using first and second neural networks, said first neural network having N input nodes and output nodes, said method comprising the steps of: pretraining said first neural network to recognize a starting point and an ending point of said periodic characteristic waveform by producing outputs which are a representation of an index number of a portion of the sample of the signal corresponding to the starting point and the ending point; transmitting said signal to a buffer having greater than N storage locations; transmitting N consecutive samples of said signal to N input nodes of said first neural network; and using the second neural network to classify said periodic characteristic waveform between said starting and ending points.

Claim 23

Independent Claim 128 words

A method of analyzing a periodic characteristic waveform in a signal using first and second neural networks, said first neural network having N input nodes and output nodes, said method comprising the steps of: pretraining said first neural network to recognize a starting point and an ending point of said periodic characteristic waveform by producing outputs which are a representation of an index number of a portion of the sample of the signal corresponding to the starting point and the ending point; transmitting said signal to a buffer having greater than N storage locations; transmitting N consecutive samples of said signal to N input nodes of said first neural network; and using the second neural network to classify said periodic characteristic waveform between said starting and ending points.

Annuity Stage	Window Opens	Payment Deadline	Surcharge Period Until	Estimated Fee	Status
3.5 Years (Year 4)	2000-01-28	2000-07-29	2001-01-28	$2000.00	Window Closed / Lapsed
7.5 Years (Year 8)	2004-01-28	2004-07-29	2005-01-28	$3760.00	Window Closed / Lapsed
11.5 Years (Year 12)	2008-01-28	2008-07-29	2009-01-28	$7700.00	Window Closed / Lapsed

Real-time waveform analysis using artificial neural networks

Abstract

Inventors

Bibliographic Details

Claims Analysis

References Cited (Backward Citations)

USPTO Annuity & Maintenance Windows

Assigned Categories & Tags

Assign Category Tag