IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0844650
(2004-05-12)
|
등록번호 |
US-7276031
(2007-10-02)
|
발명자
/ 주소 |
- Norman,Robert G.
- Ayappa,Indu A.
- Rapoport,David M.
|
출원인 / 주소 |
|
대리인 / 주소 |
Fay Kaplun & Marcia, LLP.
|
인용정보 |
피인용 횟수 :
18 인용 특허 :
30 |
초록
▼
Described is a method and system for analyzing a patient's breaths. The arrangement may include a sensor and a processor. The sensor detects data corresponding to a patient's breathing patterns over a plurality of breaths. The processor separates the detected data into data segments corresponding to
Described is a method and system for analyzing a patient's breaths. The arrangement may include a sensor and a processor. The sensor detects data corresponding to a patient's breathing patterns over a plurality of breaths. The processor separates the detected data into data segments corresponding to individual breaths. Then, the processor analyzes the data segments using a pretrained artificial neural network to classify the breaths based on a likelihood that individual ones of the breaths include an abnormal flow limitation.
대표청구항
▼
What is claimed is: 1. An arrangement for analyzing a patient's breaths, comprising: a sensor detecting data corresponding to a patient's breathing patterns over a plurality of breaths; and a processor separating the detected data into data segments, each segment corresponding to an individual brea
What is claimed is: 1. An arrangement for analyzing a patient's breaths, comprising: a sensor detecting data corresponding to a patient's breathing patterns over a plurality of breaths; and a processor separating the detected data into data segments, each segment corresponding to an individual breath, the processor analyzing the data segments using a pretrained artificial neural network to classify each breath based on a likelihood that the breath includes an abnormal flow limitation, wherein the classifications include: (i) a flow limitation is present; (ii) a flow limitation is probably present; (iii) a flow limitation is not probably present; and (iv) a flow limitation is not present. 2. The arrangement according to claim 1, wherein the abnormal flow limitation includes a predefined change in the shape of an inspiratory flow curve with respect to time. 3. The arrangement according to claim 1, wherein the neural network is trained based on at least one set of input data and cone sponding output data, the output data being generated as a function of the input data and independently of the neural network. 4. The arrangement according to claim 3, further comprising: a flow generator providing a flow of air to the patient, the flow generator being controlled by the processor. 5. The arrangement according to claim 4, wherein, upon detecting the likelihood that an individual breath includes the flow limitation, the processor generates signals transmitted to the flaw generator to adjust the flow of air, the signals being generated by the processor as a function of output data. 6. The arrangement according to claim 1, further comprising: a venting arrangement allowing gases exhaled by the patient to be diverted from an incoming flow of air. 7. The arrangement according to claim 1, wherein the neural network is provided with first input data to generate first output data, the first output data being compared to second output data to generate an error margin value, the second output data being generated independently of the neural network. 8. The arrangement according to claim 7, wherein until the error margin value is greater than a predetermined error margin, the neural network is further trained with at least one set of training input and output data. 9. The arrangement according to claim 1, wherein each data segment is further broken into a set of flow points which characterizes a period of a patient's inspiration, and wherein if the set of flow points has a substantially sinusoidal pattern, the corresponding breath is classified as being free of the flow limitation. 10. The arrangement according to claim 1, wherein each data segment includes a first time period T1 of a patient's inspiration and a second time period T2 of a patient's expiration, wherein an indicator value ID is calculated according to the following formula: ID=T1/(T1+T2), and wherein if the indicator value is above a predetennined value, the corresponding breath is classified as having the flow limitation. 11. The arrangement according to claim 1, wherein each data segment includes a first area RA of a patient's inspiration under a curve representing a patient's inspiration above a critical threshold and a second area RB under a curve representing the patient's inspiration below the critical threshold, wherein an indicator value ID is calculated according to the following formula: ID=RA/RB, and wherein, if the ID is at least equal to a predetermined value, the corresponding breath is classified as being free of the flow limitation. 12. The arrangement according to claim 1, wherein the classifications further include (v) an indeterminate classification in which the neural network is unable the classify the breath. 13. An arrangement, comprising: a sensor detecting data corresponding to a patient's breathing patterns over a plurality of breaths; an input device receiving control data corresponding to a desired diagnosis as to whether (i) a flow limitation is present; (ii) a flow limitation is probably present; (iii) a flow limitation is not probably present; and (iv) a flow limitation is not present in at least one of the breaths; and a processor coupled to the sensor and the input device, the processor separating the detected data into data segments, each segment corresponding to an individual breath, the processor running an artificial neural network and processing the data segments and the corresponding control data to refine the neural network which processes the data segments to generate output data approximating the desired diagnoses. 14. A method for analyzing a patient's breaths, comprising the steps of: obtaining data corresponding to a patient's breathing patterns over a plurality of breaths of the patient; processing the detected data into data segments, each segment corresponding to an individual breath; and analyzing each data segment using a pretrained artificial neural network to classify each breath based on a likelihood that the breath includes an abnormal flow limitation, wherein the classifications include: (i) a flow limitation is present; (ii) a flow limitation is probably present; (iii) a flow limitation is not probably present; and (iv) a flow limitation is not present. 15. The method according to claim 14, further comprising the step of: training the neural network based on at least one set of input data and corresponding output data, the output data being generated as a function of the input data and independently of the neural network. 16. The method according to claim 15, further comprising the step of: providing to the patient a flow of air generated by a flow generator. 17. The method according to claim 16, flurther comprising the steps of: upon detecting presence of the flow limitations in the individual breath, generating signals to adjust the flow of air as a function of the output data; and transmitting the signals to the flow generator. 18. The method according to claim 14, further comprising the step of: allowing, with a venting arrangement, gases exhaled by the patient to be diverted from an incoming flow of air. 19. The method according to claim 14, further comprising the steps of: testing the neural ndwork by providing first input data and collecting first output data; and comparing the first output data to second output data to generate an error margin value, wherein the second control output data is generated independently of the neural network. 20. The method according to claim 19, further comprising the step of: until the error margin value is greater than a predetermined error margin, further training the neural network using at least one set of training input and output data. 21. The method according to claim 14, further comprising the steps of: further breaking each data segment into a set of flow points which characterizes a period of a patient's inspiration; and if the set of flow points has a substantially sinusoidal patent, generating the output data indicative of absence of the flow limitation in the corresponding individual breath. 22. The method according to claim 14, further comprising the steps of: calculating a first time period T1 of a patient's inspiration and a second time period T2 of a patient's expiration in each data segment; determining an indicator value ID according to the following formula: ID=T1/(T1+T2); and generating the output data indicative of the presence of the flow limitation in the corresponding individual breath if the indicator value is above a predetermined value. 23. The method according to claim 14, further comprising the steps of: calculating, in each data segment, a first area RA of a patient's inspiration under a curve representing a patient's inspiration above a critical threshold and a second area RB under a curve representing the patient's inspiration below the critical threshold; determining an indicator value ID) according to the following formula: ID=RA/RB; and if the ID is at least equal to a predetermined value, generating the output data indicative of absence of the flow limitations in the corresponding individual breath. 24. A method, comprising the steps of: detecting, with a sensor, data corresponding a patient's breathing patterns over a plurality of breaths of the patient; receiving, with an input device, control data corresponding to a desired diagnosis as to whether (i) a flow limitation is present; (ii) a flow limitation is probably present; (iii) a flow limitation is not probably present; and (iv) a flow limitation is not present in at least one of the breaths; separating, with a processor, the detected data into data segments, each segment corresponding to an individual breath, the processor running an artificial neural network; and processing, with the processor, the data segments and the corresponding control data to refine the neural network which processes the data segments to generate output data approximating the desired diagnoses.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.