A bowel sound detection system

Abstract

The present invention relates to a bowel sound detection system, belonging to the technical field of bowel sound detection, and solves the problem of inaccurate bowel sound recognition in the prior art. The system includes: a signal acquisition device, including a laser, a distributed optical fiber sensor, and a collector; the laser is used to periodically emit laser to the distributed optical fiber sensor; the collector is used to collect the signal returned by the optical fiber sensor and send it to the bowel sound calculation module; the bowel sound calculation module is used to calculate the collected signal based on optical frequency domain reflectometry, judge whether a bowel sound occurs, and if so, determine the occurrence position of the bowel sound and extract the bowel sound signal; the bowel sound classification module is used to obtain the individual bowel sound type based on the occurrence position of the bowel sound and the bowel sound signal according to the trained bowel sound classification model. Accurate recognition of bowel sounds is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of bowel sound detection, and in particular to a bowel sound detection system. Background Art

[0002] Bowel sounds are generated by the friction between the intestinal contents (solids, gases or liquids) and the intestine or between the intestinal contents under the premise of intestinal peristalsis, which can reflect the intestinal movement state and is related to clinical applications such as the physiological state of the human body and disease diagnosis.

[0003] Currently, clinical diagnosis mainly relies on manual auscultation of bowel sounds, which has defects such as strong subjectivity, few evaluation parameters, high missed diagnosis rate, inability to record data and non-repeatability. At the same time, the auscultation quality is limited by the experience and ability of the auscultator. From the perspective of informatics, bowel sounds are a kind of vibration signal, containing rich disease-related information, such as the main frequency, loudness, etc. These information are difficult or impossible to be directly recognized by the human ear. Compared with other physiological sounds of the human body (such as heart sounds and breath sounds), bowel sounds have poor periodicity, low regularity, strong timeliness and many influencing factors, bringing many difficulties to clinical applications. Therefore, the accuracy and clinical value of manual auscultation of bowel sounds face great challenges.

[0004] Existing bowel sound acquisition devices usually adopt traditional acoustic sensors such as coil microphones, and fix one or more sensors near the abdominal cavity through a belt to achieve the monitoring of bowel sounds. The belt-type acoustic sensor array is in contact with the human body, and a large amount of friction noise will be generated during human activities, bringing serious interference to the bowel sound signal. At the same time, the noise signal will greatly increase the data volume and reduce the data quality, making it difficult for subsequent bowel sound identification. The combination of the belt-type sensing array and the traditional acoustic probe limits the number of sensors, making the accuracy of the time-delay method acoustic positioning insufficient; since the wavelength of sound waves is much larger than the abdominal cavity size, and the reflection and superposition of sound waves in human tissues are relatively serious, the accuracy of positioning the bowel sound position by traditional methods is poor. In addition, due to human differences and movements during wearing, problems such as inaccurate initial position positioning and displacement of the sensor will occur, resulting in the failure of bowel sound positioning and the deterioration of data quality. Therefore, there are many deficiencies in the basic stage of data acquisition for the current monitoring and identification of bowel sounds, resulting in a significant increase in the difficulty of subsequent bowel sound recognition. Summary of the Invention

[0005] In view of the above analysis, an embodiment of the present invention aims to provide a bowel sound detection system to solve the problem of inaccurate recognition of existing bowel sounds.

[0006] On the one hand, an embodiment of the present invention provides a bowel sound detection system, which includes:

[0007] A signal acquisition device, comprising a laser, a distributed optical fiber sensor, and a collector; the laser is used to periodically emit laser light to the distributed optical fiber sensor; the collector is used to collect the signal returned by the optical fiber sensor and send it to the bowel sound calculation module;

[0008] The bowel sound calculation module is used to calculate the collected signal based on optical frequency domain reflectometry, determine whether bowel sounds occur, and if so, determine the occurrence location of the bowel sounds and extract the bowel sound signal;

[0009] The bowel sound classification module is used to obtain the bowel sound type of an individual based on the occurrence location of the bowel sounds and the bowel sound signal according to the trained bowel sound classification model.

[0010] Based on a further improvement of the above technical solution, the distributed optical fiber sensor includes a test optical fiber, a reference optical fiber, a first coupler, and a circulator; the first coupler is connected to the laser and is used to divide the light source into two optical paths, one path is injected into the reference optical fiber, and the other path is injected into the test optical fiber through the circulator;

[0011] The collector includes a second coupler and a photodetector. The second coupler is connected to the reference optical fiber and the circulator and is used to couple the signals returned by the reference optical fiber and the test optical fiber and send them to the photodetector; the photodetector converts the optical signal into an electrical signal and sends it to the bowel sound calculation module.

[0012] Based on a further improvement of the above technical solution, the signal acquisition device further includes a cylindrical housing; there is an annular groove on the inner wall of the cylindrical housing; the test optical fiber of the distributed optical fiber sensor is wound on the inner wall of the cylindrical housing, and the test optical fiber of the distributed optical fiber sensor is embedded in the groove on the inner wall of the cylindrical housing.

[0013] Based on a further improvement of the above technical solution, the bowel sound calculation module calculates the collected signal in the following manner to determine whether bowel sounds occur, and if so, determines the occurrence location of the bowel sounds:

[0014] Obtain a reference signal; the reference signal has the same length as the collected signal;

[0015] Based on the similarity between the reference signal and the collected signal, determine whether bowel sounds occur, and if so, locate the occurrence location of the bowel sounds based on the similarity.

[0016] Based on a further improvement of the above technical solution, the following method is used to determine whether bowel sounds occur based on the similarity between the reference signal and the collected signal:

[0017] Segment the reference signal and the collected signal respectively according to the emission period of the laser;

[0018] Convert each segmented signal of the reference signal and the acquired signal to the range domain through fast Fourier transform;

[0019] For each range-domain segment of the acquired signal, use a sliding window to divide the range-domain segment of the acquired signal and the corresponding range-domain segment of the reference signal into multiple sub-segments respectively;

[0020] For each sub-segment of the range-domain segment, use inverse Fourier transform to convert the sub-segment and the corresponding sub-segment of the reference signal into time-domain signals respectively; calculate the cross-correlation graph of the two groups of time-domain signals to obtain the similarity of the sub-segment;

[0021] If there is a sub-segment with a similarity less than the first threshold in the range segment, it is determined that bowel sounds have occurred.

[0022] Based on a further improvement of the above technical solution, the bowel sound classification model includes:

[0023] A convolutional encoding module, used to extract features from the bowel sound signal to obtain a feature frame sequence; randomly mask the feature frame sequence and input it into the BERT encoding module;

[0024] A BERT encoding module, used to output the hidden representation of each feature frame based on the randomly masked feature frame sequence;

[0025] A classification module, used to predict the type of bowel sounds based on the occurrence location of bowel sounds and the hidden representation of each feature frame.

[0026] Based on a further improvement of the above technical solution, the following method is used to obtain a trained bowel sound classification model:

[0027] Obtain multiple bowel sound signals to construct a training sample set;

[0028] Combine the convolutional encoding module and the BERT encoding module to form a pre-training model; add a clustering module to the pre-training model; the clustering module is used to cluster the signal frames of the bowel sound signal to obtain the labels of each clustering type; the BERT encoding module also outputs the prediction results of the clustering type of each feature frame;

[0029] Based on the labels of the clustering types of each feature frame of the sample and the predicted clustering types, calculate the loss to train the pre-training model to obtain a trained pre-training model;

[0030] Fix the parameters of the convolutional encoding module, and fine-tune the bowel sound classification model based on the training sample set to obtain a trained bowel sound classification model.

[0031] Based on a further improvement of the above technical solution, the following formula is used to calculate the loss:

[0032] ;

[0033] Among them, represents the loss of the masked feature frames in the sample, represents the loss of the unmasked feature frames in the sample, represents the weight.

[0034] Based on the further improvement of the above technical solution, the following formula is used to calculate the loss of the masked feature frames:

[0035] ;

[0036] Among them, represents the clustering type loss of the i-th masked feature frame, represents the distance loss of the i-th masked feature frame, represents the number of masked feature frames in the sample.

[0037] Based on the further improvement of the above technical solution, the following formula is used to calculate the distance loss of the i-th masked feature frame:

[0038] ;

[0039] Among them, represents the hidden representation of the i-th masked feature frame, represents the hidden representation of the p-th feature frame, represents the hidden representation of the q-th feature frame; the clustering type label of the p-th feature frame is the same as that of the i-th masked feature frame, and the clustering type label of the q-th feature frame is different from that of the i-th masked feature frame, represents the margin parameter.

[0040] Compared with the prior art, the present invention collects signals in a non-contact manner through a signal acquisition device, judges and calculates bowel sounds through a bowel sound calculation module, so as to accurately calculate the occurrence position and sound signal, and predicts the type based on a trained bowel sound classification model through a bowel sound classification module, so as to accurately identify bowel sounds.

[0041] In the present invention, the above technical solutions can also be combined with each other to achieve more preferred combination schemes. Other features and advantages of the present invention will be described in the subsequent specification, and some advantages can be made obvious from the specification, or understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained through the content specifically pointed out in the specification and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings are only for the purpose of showing specific embodiments and are not considered as a limitation to the present invention. Throughout the drawings, the same reference signs denote the same components;

[0043] Figure 1 It is a block diagram of the bowel sound detection system according to an embodiment of the present invention;

[0044] Figure 2 It is a schematic structural diagram of a part of the signal acquisition device according to an embodiment of the present invention;

[0045] Figure 3 It is a schematic diagram of the use of the signal acquisition device according to an embodiment of the present invention;

[0046] Reference signs:

[0047] 1 - The outer shell of the cylindrical shell, 2 - Optical fiber, 3 - Movable bed, 4 - Sound insulation material, 5 - Bell mouth, 6 - The inner shell of the cylindrical shell, 7 - Optical fiber cross-section, 8 - Damping cavity. Specific embodiments

[0048] The following will specifically describe the preferred embodiments of the present invention in conjunction with the accompanying drawings. Among them, the accompanying drawings form a part of this application and are used together with the embodiments of the present invention to explain the principle of the present invention, and are not used to limit the scope of the present invention.

[0049] Optical fiber is a new type of communication medium. As a transmission carrier of signals, its transmission distance ranges from a few meters to dozens or even hundreds of kilometers. In many fields, it has gradually replaced traditional wireless communication technologies. When light transmits in an optical fiber, photons collide with matter molecules to generate scattering. Distributed optical fiber sensing realizes distributed optical fiber sensing by detecting the one-to-one correspondence between these scattering signals and external physical quantities.

[0050] Any position on the test optical fiber of the distributed optical fiber sensor can be used as a sensor. Therefore, the number of measurement points is much higher than that of traditional sensor acquisition methods. Moreover, the distributed optical fiber sensor does not need to contact the measured object, so the positioning is more accurate and the quality of the collected signals is higher, thus improving the accuracy of detection.

[0051] Based on this, a specific embodiment of the present invention discloses a bowel sound detection system, as Figure 1 shown, including:

[0052] A signal acquisition device, including a laser, a distributed optical fiber sensor, and a collector; the laser is used to periodically emit laser light to the distributed optical fiber sensor; the collector is used to collect the signals returned by the optical fiber sensor and send them to the bowel sound calculation module;

[0053] The bowel sound calculation module is used to calculate the acquired signal based on optical frequency domain reflection, determine whether bowel sounds occur, and if so, determine the occurrence location of the bowel sounds and extract the bowel sound signal.

[0054] The bowel sound classification module is used to obtain the bowel sound type of an individual based on the occurrence location of the bowel sounds and the bowel sound signal using a trained bowel sound classification model.

[0055] Compared with the prior art, the bowel sound detection system provided in this embodiment collects signals in a non-contact manner through a signal acquisition device, performs bowel sound judgment and calculation through the bowel sound calculation module, so as to accurately calculate the occurrence location and sound signal, and performs type prediction through the bowel sound classification module based on a trained bowel sound classification model, so as to accurately identify bowel sounds.

[0056] During implementation, the light source of the laser is a tunable laser source, and the laser emits a linearly chirped signal in each emission cycle.

[0057] The distributed fiber optic sensor includes a test optical fiber, a reference optical fiber, a first coupler, and a circulator; the first coupler is connected to the laser and is used to divide the light source into two optical paths, one is injected into the reference optical fiber, and the other is injected into the test optical fiber through the circulator;

[0058] The collector includes a second coupler and a photodetector. The second coupler is connected to the reference optical fiber and the circulator and is used to couple the signals returned by the reference optical fiber and the test optical fiber and send them to the photodetector; the photodetector converts the optical signal into an electrical signal and sends it to the bowel sound calculation module.

[0059] The light source is divided into two beams by the first coupler, one beam enters the reference optical fiber; the other beam enters the circulator through port 1 of the circulator, exits through port 2 and enters the test optical fiber. Part of the light in the test optical fiber will be reflected back. The reflected light enters the second coupler through port 3 of the circulator. The optical paths of the two beams of light are different and the optical frequencies are different. Therefore, a beat frequency signal is generated at the output of the second coupler. After the photodetector collects the beat frequency signal, it is converted into an electrical signal and sent to the bowel sound calculation module.

[0060] During implementation, in order to facilitate signal acquisition, the signal acquisition device further includes a cylindrical housing; there is an annular groove on the inner wall of the cylindrical housing; the test optical fiber of the distributed fiber optic sensor is wound on the inner wall of the cylindrical housing, and the test optical fiber of the distributed fiber optic sensor is embedded in the groove on the inner wall of the cylindrical housing.

[0061] During implementation, annular grooves are arranged at equal intervals on the inner wall of the cylindrical housing.

[0062] During implementation, as Figure 2As shown, the cylindrical shell includes an outer shell and an inner wall. Sound insulation materials are filled between the outer shell and the inner wall. The test optical fiber is wrapped in the shock absorption cavity, thereby reducing the influence of the outside world on the sensor. In order to improve the detection sensitivity, the shock absorption cavity on the measurement surface of the optical fiber uses a thin shell or no shell, so that the optical fiber can detect the physiological sound signal more accurately. By setting grooves on the inner wall, the test optical fiber is embedded in the grooves. The cross-section of the grooves is in the shape of a horn, forming a horn mouth. The configuration of the horn mouth can enable the sensor to specifically monitor a specific small area of the abdomen, thereby avoiding environmental interference and vibration interference from other parts (such as breathing, heartbeat, etc.), and focusing on the sound signal collection in the small area corresponding to the horn mouth, realizing physical noise reduction, and being more suitable for receiving bowel sound signals.

[0063] During implementation, the individual to be detected lies on a mobile hospital bed, and the mobile hospital bed is pushed into the cylindrical shell for easy detection, as Figure 3 shown. The position corresponding to the abdomen on the hospital bed is a hollow structure, so that the patient's back directly faces the lower tactile sensor array, so as not to affect the propagation of bowel sound signals.

[0064] During implementation, the signal acquisition device performs signal acquisition for 2 to 4 minutes and sends the acquired signal to the bowel sound calculation module.

[0065] During implementation, after receiving the acquired signal, the bowel sound calculation module calculates the acquired signal in the following manner to determine whether bowel sounds occur and determine the location where the bowel sounds occur:

[0066] S21. Obtain a reference signal; the reference signal has the same length as the acquired signal;

[0067] S22. Determine whether bowel sounds occur based on the similarity between the reference signal and the acquired signal. If so, locate the location where the bowel sounds occur based on the similarity.

[0068] During implementation, the reference signal is the signal acquired by the signal acquisition device when there is no sound, and the reference signal has the same length as the signal acquired during detection. It is determined whether bowel sounds occur based on the similarity between the reference signal and the acquired signal.

[0069] Specifically, the following method is used to determine whether bowel sounds occur based on the similarity between the reference signal and the acquired signal:

[0070] S221. Segment the reference signal and the acquired signal respectively according to the emission period of the laser;

[0071] S222. Convert each segmented signal of the reference signal and the acquired signal to the distance domain through fast Fourier transform;

[0072] S223. For each distance-domain segment of the acquired signal, use a sliding window to divide the distance-domain segment of the acquired signal and the corresponding distance-domain segment of the reference signal into multiple sub-segments respectively;

[0073] S224. For each sub-segment of the distance-domain segment, use the inverse Fourier transform to transform the sub-segment and the corresponding sub-segment of the reference signal into time-domain signals respectively; calculate the cross-correlation graph of the two groups of time-domain signals to obtain the similarity of the sub-segment;

[0074] S225. If there is a sub-segment with a similarity less than the first threshold in the distance segment, it is determined that bowel sounds have occurred.

[0075] During implementation, first, according to the emission period of the laser, segment the reference signal and the acquired signal respectively, and each segment corresponds to the signal acquired by one laser detection. The segmentation of the reference signal and the segmentation of the acquired signal correspond one by one.

[0076] For each group of segments (the i-th segment of the acquired signal and the i-th segment of the reference signal ), use the fast Fourier transform to transform each segmented signal of the reference signal and the acquired signal into the frequency domain, multiply the frequency-domain signal by a coefficient and convert it into the distance-domain signal of the test optical fiber proportionally, that is, obtain the corresponding distance-domain segment and .

[0077] Use a sliding window to divide the i-th distance-domain segment and into multiple sub-segments respectively. During implementation, the sliding step of the sliding window can be less than the window length, that is, there can be an overlap between two adjacent windows. The sub-segments of and

[0078] are in one-to-one correspondence.

[0079] Each sub-segment corresponds to the corresponding position of the test optical fiber. and For each sub-segment, the j-th sub-segment

[0080] During implementation, use the following method to calculate the cross-correlation graph of the two groups of time-domain signals to obtain the similarity of the sub-segment:

[0081] Use a weighted cross-correlation function to perform cross-correlation processing on the two groups of time-domain signals to obtain the cross-correlation graph;

[0082] According to the formula Obtain the similarity of the sub-segments, where represents the number of points in the cross-correlation graph whose amplitude exceeds the second threshold, represents the total number of points in the cross-correlation graph. During implementation, set the first threshold and the second threshold according to the requirements of detection sensitivity and false alarm rate.

[0083] If there is a sub-segment with a similarity less than the first threshold in the i-th distance segment, it is considered that bowel sounds occur at the acquisition moment corresponding to the i-th distance segment, and the position corresponding to the sub-segment with the minimum similarity is the position where the bowel sounds occur, thereby determining the bowel sound localization.

[0084] During implementation, the collected signal can be demodulated through a phase demodulation algorithm to obtain the bowel sound signal.

[0085] After obtaining the bowel sound signal, input the occurrence position and the bowel sound signal into the trained bowel sound classification model to obtain the bowel sound type of the individual.

[0086] It should be noted that the bowel sound types include hyperactive, active, normal, hypoactive, and absent.

[0087] Specifically, the bowel sound classification model includes:

[0088] A convolutional encoding module, which is used to extract features from the bowel sound signal to obtain a feature frame sequence; randomly mask the feature frame sequence and then input it into the BERT encoding module;

[0089] A BERT encoding module, which is used to output the hidden representation of each feature frame based on the randomly masked feature frame sequence;

[0090] A classification module, which is used to predict the bowel sound type based on the occurrence position of the bowel sound and the hidden representation of each feature frame.

[0091] During implementation, the classification module can adopt an existing classifier.

[0092] During implementation, the following method is adopted to obtain the trained bowel sound classification model:

[0093] Obtain multiple bowel sound signals to construct a training sample set;

[0094] Form a pre-training model with the convolutional encoding module and the BERT encoding module; add a clustering module to the pre-training model; the clustering module is used to cluster the signal frames of the bowel sound signal to obtain the labels of each clustering type; the BERT encoding module also outputs the predicted results of the clustering type of each feature frame;

[0095] Based on the labels of the clustering type of each feature frame of the sample and the predicted clustering type, calculate the loss to train the pre-training model to obtain the trained pre-training model;

[0096] Fix the parameters of the convolutional encoding module and fine-tune the bowel sound classification model based on the training sample set to obtain a trained bowel sound classification model.

[0097] During implementation, multiple bowel sound signals are obtained to construct a training sample set. Since the efficiency of manual annotation is low, only some of the bowel sound signals need to have corresponding bowel sound type labels. By pre-training the unlabeled samples and then fine-tuning on the labeled samples, the performance and efficiency of the model can be improved.

[0098] During implementation, the sound signal is a continuous time series signal. To extract signal features more deeply and improve the detection accuracy, the bowel sound classification model extracts features from the bowel sound signal through the convolutional encoding module and divides the bowel sound signal into multiple consecutive feature frames. For example, when extracting features from a bowel sound signal with a sampling rate of 16 kHz, a feature sequence with a frame rate of 20 ms is obtained. During implementation, the convolutional encoding module can adopt an existing convolutional neural network module.

[0099] During implementation, according to the frame rate obtained by the convolutional encoding module, that is, the length of each feature frame, the original bowel sound signal is frame-segmented to obtain the signal frames of the original bowel sound signal, and the signal frames and feature frames correspond one by one. The pre-training is based on unlabeled samples. To train the model, a clustering module is added to the pre-trained model to cluster each signal frame and add clustering labels, thereby training the model.

[0100] During implementation, the acoustic features (such as MFCC) of the original signal frame can be used as the input of the clustering module. After clustering, the centroid point or center point of each clustering type is used as the label of the clustering type.

[0101] During implementation, the obtained feature frame sequence is randomly masked and then input into the BERT encoding module. The BERT encoding module predicts and outputs the hidden representation of each feature frame based on the randomly masked feature frame sequence, and also outputs the prediction result of the clustering type of each feature frame.

[0102] During implementation, the BERT encoding module adopts an existing BERT encoder structure, projects the hidden representation corresponding to each feature frame output by the BERT encoder into the space corresponding to the clustering label through a projection matrix, that is, obtains the prediction result of the clustering type corresponding to the feature frame, calculates the loss based on the clustering type label and prediction result of the feature frame, and updates the parameters of the pre-trained model.

[0103] Specifically, the following formula is used to calculate the loss:

[0104] ;

[0105] where Represents the loss of the masked feature frames in the sample, Represents the loss of the unmasked feature frames in the sample, Represents the weight, .

[0106] During implementation, due to random masking, the losses for masked and unmasked feature frames are calculated separately.

[0107] Specifically, the following formula is used to calculate the loss of the masked feature frames:

[0108] ;

[0109] Where, Represents the clustering type loss of the i-th masked feature frame, Represents the distance loss of the i-th masked feature frame, Represents the number of masked feature frames in the sample.

[0110] Specifically, the following formula is used to calculate the clustering type loss of the i-th masked feature frame:

[0111] ;

[0112] Where, Represents the hidden representation of the i-th masked feature frame, Represents the label of the clustering type of the i-th masked feature frame, C represents the number of clustering types, Represents the scaling parameter, A represents the projection matrix, Represents the label of the j-th clustering type.

[0113] The label of the clustering type of the i-th masked feature frame, that is, the clustering type label of the corresponding signal frame.

[0114] For the i-th masked feature frame, it should be close in distance to feature frames of the same type and far from feature frames of different types. Therefore, the following formula is used to calculate the distance loss of the i-th masked feature frame:

[0115] ;

[0116] Where, Represents the hidden representation of the i-th masked feature frame, Represents the hidden representation of the p-th feature frame, Represents the hidden representation of the q-th feature frame; the clustering type label of the p-th feature frame is the same as that of the i-th masked feature frame, and the clustering type label of the q-th feature frame is different from that of the i-th masked feature frame, Represents the margin parameter.

[0117] Used to represent the minimum distance difference between feature frames of different clustering types.

[0118] By considering the label loss and clustering loss, the classification accuracy of the model is improved.

[0119] Calculate the loss of the unmasked feature frames in the same way , and update the parameters of the pre-trained model based on the loss in reverse. When the training stop condition is reached, stop the training to obtain the trained pre-trained model. The training stop condition can be, for example, reaching the preset loss accuracy or the number of training times.

[0120] After obtaining the trained pre-trained model, the convolutional encoding module, the BERT encoding module, and the classification module form a bowel sound classification model. Fix the parameters of the convolutional encoding module and fine-tune the bowel sound classification model based on the training sample set to obtain the trained bowel sound classification model. Among them, remove the projection matrix of the BERT encoding module and replace it with a randomly initialized softmax layer. The classification module predicts the type of bowel sound signal based on the occurrence position of the bowel sound and the hidden representation of each feature frame.

[0121] During fine-tuning, the classification loss can be calculated through the cross-entropy loss function to update the parameters of the BERT encoding module and the classification module. When the training stop condition is reached, stop the training to obtain the trained bowel sound classification model.

[0122] The bowel sound classification module can quickly and accurately obtain the bowel sound type of an individual by inputting the occurrence position of the bowel sound and the bowel sound signal into the trained bowel sound classification model.

[0123] Those skilled in the art can understand that all or part of the processes of implementing the above embodiment methods can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a disk, an optical disc, a read-only memory, or a random access memory, etc.

[0124] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A bowel sound detection system, characterized in that, The system includes: A signal acquisition device, including a laser, a distributed optical fiber sensor, and a collector; the laser is used to periodically emit laser light to the distributed optical fiber sensor; the collector is used to collect the signals returned by the optical fiber sensor and send them to the bowel sound calculation module; A bowel sound calculation module, which is used to calculate the collected signals based on optical frequency domain reflection, judge whether bowel sounds occur, and if so, determine the occurrence position of the bowel sounds and extract the bowel sound signals; A bowel sound classification module, which is used to obtain the bowel sound type of an individual based on the occurrence position of the bowel sounds and the bowel sound signals according to the trained bowel sound classification model; The distributed optical fiber sensor includes a test optical fiber, a reference optical fiber, a first coupler, and a circulator; the first coupler is connected to the laser and is used to divide the light source into two optical paths, one is injected into the reference optical fiber, and the other is injected into the test optical fiber through the circulator; The collector includes a second coupler and a photodetector. The second coupler is connected to the reference optical fiber and the circulator and is used to couple the signals returned by the reference optical fiber and the test optical fiber and send them to the photodetector; the photodetector converts the optical signal into an electrical signal and sends it to the bowel sound calculation module; The signal acquisition device further includes a cylindrical housing; the cylindrical housing includes an outer shell and an inner wall, and a sound insulation material is filled between the outer shell and the inner wall. The test optical fiber is wrapped in a shock absorption cavity; there is an annular groove on the inner wall of the cylindrical housing, and the cross section of the groove is in the shape of a horn, forming a horn mouth; the test optical fiber of the distributed optical fiber sensor is wound on the inner wall of the cylindrical housing, and the test optical fiber of the distributed optical fiber sensor is embedded in the groove on the inner wall of the cylindrical housing.

2. The bowel sound detection system according to claim 1, characterized in that The bowel sound calculation module calculates the collected signals in the following way to judge whether bowel sounds occur, and if so, determine the occurrence position of the bowel sounds: Obtain a reference signal; the reference signal has the same length as the collected signal; Judge whether bowel sounds occur based on the similarity between the reference signal and the collected signal. If so, locate the occurrence position of the bowel sounds based on the similarity.

3. The bowel sound detection system according to claim 2, wherein The following method is used to judge whether bowel sounds occur based on the similarity between the reference signal and the collected signal: Segment the reference signal and the collected signal respectively according to the emission period of the laser; Convert each segmented signal of the reference signal and the collected signal to the distance domain through fast Fourier transform; For each distance domain segment of the collected signal, use a sliding window to divide the distance domain segment of the collected signal and the corresponding distance domain segment of the reference signal into multiple sub-segments respectively; For each sub-segment of this distance domain segment, use inverse Fourier transform to convert this sub-segment and the corresponding sub-segment of the reference signal into time domain signals respectively; calculate the cross-correlation graph of the two groups of time domain signals to obtain the similarity of this sub-segment; If there is a sub-segment with a similarity less than the first threshold in this distance domain segment, it is judged that bowel sounds have occurred; The position corresponding to the sub-segment with the smallest similarity is the occurrence position of the bowel sounds.

4. The bowel sound detection system according to claim 1, wherein The bowel sound classification model includes: A convolutional coding module, which is used to extract features from the bowel sound signals to obtain a feature frame sequence; randomly mask the feature frame sequence and then input it into the BERT coding module; The BERT encoding module is used to output the hidden representation of each feature frame based on the sequence of feature frames after random masking; The classification module is used to predict the type of bowel sound based on the occurrence location of the bowel sound and the hidden representation of each feature frame.

5. The bowel sound detection system according to claim 4, characterized in that, The trained bowel sound classification model is obtained in the following manner: Obtain multiple bowel sound signals to construct a training sample set; The convolutional encoding module and the BERT encoding module are combined to form a pre-training model; a clustering module is added to the pre-training model; the clustering module is used to cluster the signal frames of the bowel sound signal to obtain the labels of each clustering type; the BERT encoding module also outputs the prediction results of the clustering type of each feature frame; Based on the labels of the clustering type of each feature frame of the sample and the predicted clustering type, calculate the loss to train the pre-training model to obtain a trained pre-training model; Fix the parameters of the convolutional encoding module and fine-tune the bowel sound classification model based on the training sample set to obtain a trained bowel sound classification model.

6. The bowel sound detection system according to claim 5, wherein The following formula is used to calculate the loss: ； Among them, represents the loss of the masked feature frames in the sample, represents the loss of the unmasked feature frames in the sample, represents the weight, and the loss of the unmasked feature frames is calculated in the same way as the loss of the masked feature frames is calculated .

7. The bowel sound detection system according to claim 6, characterized in that The following formula is used to calculate the loss of the masked feature frame: ； Among them, represents the clustering type loss of the i-th masked feature frame, represents the distance loss of the i-th masked feature frame, represents the number of masked feature frames in the sample.

8. The bowel sound detection system according to claim 7, wherein The following formula is used to calculate the distance loss of the i-th masked feature frame: ； Among them, represents the hidden representation of the i-th masked feature frame, represents the hidden representation of the p-th feature frame, represents the hidden representation of the q-th feature frame; the clustering type label of the p-th feature frame is the same as that of the i-th masked feature frame, and the clustering type label of the q-th feature frame is different from that of the i-th masked feature frame, represents the margin parameter.

Images