Bioelectrical impedance cancer detection method based on deep learning

Abstract

The invention discloses a bioelectrical impedance cancer detection method based on deep learning, and the method comprises the following steps: S1, obtaining electrical signal data of a benign sample and a malignant sample of a cancer tissue through an electrical impedance imaging system, and combining the electrical signal data with physiological data; s2, performing global feature extraction on the electric signal by using a matrix decomposition module; s3, taking a global feature extraction result and original electric signal data as new input, and further extracting signal features by using a data dependency relationship; and S4, mapping and integrating the extracted features into a final classification result. The electrical signal data of the cancer tissue is acquired by using the electrical impedance imaging system, global feature extraction is performed through the matrix decomposition module, the result of global feature extraction and the original electrical signal data are used as new input, and the signal features are further extracted by using the data dependency relationship, so that the utilization rate of the data can be improved; and the accuracy and reliability of cancer detection are further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cancer detection, and in particular to a bioelectrical impedance cancer detection method based on deep learning. Background Art

[0002] Cancer is one of the major diseases that seriously threaten human health in the world today. Its early and accurate detection is crucial to improving patient survival rates and treatment outcomes. Traditional cancer detection methods mainly include tissue biopsy, imaging examinations, and tumor marker detection. Although tumor marker detection is relatively convenient, its specificity and accuracy are affected by many factors. A single marker often cannot achieve accurate diagnosis, and combined detection increases complexity and false positive rates.

[0003] In the field of healthcare, deep learning has been tried and applied to medical imaging diagnosis, disease prediction and other aspects, and has shown advantages over traditional methods. Introducing deep learning technology into bioimpedance cancer detection is expected to overcome the shortcomings of traditional methods in feature extraction and pattern recognition. By learning and training a large number of bioimpedance data of different individuals, different cancer types and stages, a highly accurate cancer detection model can be constructed, thereby achieving more sensitive, more accurate and non-invasive early cancer screening and diagnosis, opening up new ways for cancer prevention and treatment.

[0004] In the prior art, the Chinese patent with publication number CN117481630A discloses "a method for detecting breast cancer based on bioelectrical impedance analysis". This method uses bioelectrical impedance analysis to obtain the electrical signal data of breast tissue, detects the data through a multi-dimensional feature extraction network module, and improves the validity and richness of the data by designing a segmentation mechanism and a channel synthesis network module. However, this method ignores the temporal characteristics in the data and fails to effectively capture the dynamic changes in the tumor growth process. The detection method of bioelectrical impedance technology based on deep learning provides a new solution for cancer detection.

[0005] Therefore, a deep learning-based bioelectrical impedance cancer detection method is proposed, which can more effectively extract the electrical characteristics of cancer and overcome the limitations of existing technologies. Summary of the invention

[0006] In response to the above defects or improvement needs of the prior art, the present invention provides a bioelectrical impedance cancer detection method based on deep learning. Global feature extraction is performed on the acquired electrical signal data of cancer tissue through a matrix decomposition module. At the same time, the results of the global feature extraction are used together with the original electrical signal data as new input, and the extracted feature maps are integrated into the final classification results, which can improve data utilization and further enhance the accuracy and reliability of cancer detection.

[0007] To achieve the above object, according to one aspect of the present invention, a bioelectrical impedance cancer detection method based on deep learning is provided, and the method comprises the following steps: S1: using an electrical impedance imaging system to obtain electrical signal data of benign and malignant cancer tissue samples, and combining them with physiological data; the electrical signal data includes electrical parameters, electrical impedance data, and frequency response data; S2: extracting global features of the electrical signal using a matrix decomposition module; S3: taking the result of the global feature extraction together with the original electrical signal data as new input, and further extracting electrical signal features by using data dependency; S4: Integrate the extracted feature maps into the final classification results.

[0008] As an embodiment of the present application, step S1 specifically includes the following steps: S11: Associating different samples with physiological data, wherein the physiological data includes age, gender, personal bad habits and body mass index; S12: The electrical impedance amplitude and phase at multiple frequencies are measured and a frequency response curve is drawn to reflect the tissue's response to currents of different frequencies.

[0009] As an embodiment of the present application, step S2 specifically includes: S21: First, the input electrical signal data is processed into time series data by a 1X1 convolutional layer; S22: decomposing the electrical signal processed by the 1X1 convolutional layer into low-rank factors and basis sequences; S23: Use the temporal convolutional network to extract time-dependent features from the basis sequence, and combine it with the low-rank factor for global feature fusion to achieve global feature extraction.

[0010] As an embodiment of the present application, the calculation formula for decomposing the electrical signal processed by the 1×1 convolution layer into low-rank factors and basis sequences in step S22 is as follows:

[0011]

[0012] in, is the input electrical signal data, is a low-rank factor used to capture the global correlation between signals, is the basis sequence, which is used to reflect the temporal pattern of the signal. is the electrical signal dimension, is the number of time steps.

[0013] As an embodiment of the present application, step S23 specifically includes: S231: Base sequence The input is passed to the temporal convolutional network, and the long-term dependencies in the time dimension are captured through the extended convolution and residual connection of the temporal convolutional network. The convolution operation formula of each layer is as follows:

[0014] in, is the number of layers, is the dilation factor of the current layer, is the weight of the convolution kernel, is the output of the previous layer; The residual connection operation formula is as follows:

[0015] in, yes No. The output of the layer, is the input of the previous layer; S232: Low rank factor and the basis sequence processed by the temporal convolutional network Fusion is performed and splicing is performed on the feature dimension to form a complete global feature , the calculation formula is as follows:

[0016]

[0017] in, is the concatenated global feature, and its dimension is , represents the original electrical signal dimension, Represents the dimension of global features, represents the number of time steps, including the low-rank factor and base sequence information.

[0018] As an embodiment of the present application, step S3 specifically includes: S31: splicing and fusing the result of global feature extraction with the original electrical signal data; S32: adding position codes to the spliced ​​and fused data obtained in step S31 to obtain final input data; S33: Input the final input data obtained in step S32 into a model built based on a deep learning framework to further extract global features.

[0019] As an embodiment of the present application, the calculation formula for splicing and fusing the global features with the original electrical signal data in step S31 is as follows:

[0020] in, is the newly entered data, is the original electrical signal data, It is a global feature; Global Features And the original electrical signal data Splicing is performed on the channel dimension or time dimension, and the new input data after splicing on the time dimension , the dimension is .

[0021] As an embodiment of the present application, step S32 specifically includes: S321: Convert the input data into a suitable sequence format, concatenate according to the time step, and the number of time steps No change, feature dimension increased, new input data is ,in is the new feature dimension, indicating the length of the feature vector at each time step; S322: Use an absolute position encoder to add position information, and generate position information for each time step through a function. The position code is calculated by sine and cosine functions, and the calculation formula is as follows:

[0022]

[0023] in, is the positional encoding, is the time step, is the index of the feature dimension, is the feature dimension size, and the position encoding encodes each time step through sine and cosine functions. Encode the position Add to the new input data after fusion It is used to integrate the time location information into the features to obtain the final input data , and its calculation formula is as follows:

[0024] .

[0025] As an embodiment of the present application, step S33 specifically includes: S331: Processed input data When input into the model, the input data is first transformed into Converted into query, key and value, the calculation formula is as follows:

[0026]

[0027] in, is the learned weight matrix, They are query, key and value, is the dimension of the key, As the input of the self-attention mechanism, it is used to calculate the dependencies between each time step; S332: Calculate the similarity between the query and the key, and use this similarity to weight the value. The calculation formula is as follows:

[0028]

[0029] in, is the dimension of the key, used for scaling; Use the function to convert the attention score into a weight, and the calculation formula is as follows:

[0030] in, Representation query and key The attention weights between them and the final weighted sum value; S333: Divide the query, key, and value into multiple heads for calculation, so as to capture the feature information in different subspaces in parallel. The calculation method of each head is the same as above, and finally the outputs of all heads are concatenated. The calculation formula is as follows:

[0031]

[0032] in, is the number of heads, and the output of each head is , is the output linear transformation matrix, is the output; S334: A neural network consisting of two layers of linear transformation is used to further extract features, and each layer of linear transformation is followed by Activation function, the calculation formula of the first layer linear transformation is as follows:

[0033] in, is the weight matrix of the first layer linear transformation, is the bias term; The calculation formula of the second-layer linear transformation is as follows:

[0034] in, The weight matrix of the second layer linear transformation, is the bias term, is the output of the neural network; S335: Outputting the neural network Perform residual connection and layer normalization to get the final output .

[0035] As an embodiment of the present application, step S4 specifically includes: After processing, the final output is obtained It can be used for downstream tasks. The output formula of the classification task is as follows:

[0036] in, is the output, The layer maps the output to a probability space for classification tasks. The calculation formula of the layer is as follows:

[0037] in, Represents the model for category The output value of represents the total number of categories, In Represents a given input , the model predicts the category probability; pass Each probability value output by the layer represents the possibility that the input electrical signal belongs to each category; ultimately, the model will select the category with the highest probability as the classification result.

[0038] The beneficial effects of the present invention are: (1) The present invention uses an electrical impedance imaging system to obtain electrical signal data of cancer tissue. By measuring and analyzing the electrical signal data of cancer tissue, information about tissue electrical parameters, electrical impedance data, and frequency response data can be obtained. In combination with individual physiological data (such as age, gender, body mass index, etc.), it helps to eliminate the interference caused by individual differences, can fully reflect the electrical characteristics of the tissue, and help the model better distinguish between normal and cancerous tissues. The generalization ability of the model is enhanced by using multimodal input. Through cross-modal feature learning, the bioelectrical impedance signal and the patient's physiological characteristics can be effectively utilized to improve the reliability and accuracy of detection.

[0039] (2) The present invention decomposes the electrical signal into low-rank factors and basis sequences through a matrix decomposition module, which can effectively extract global time series features from the electrical signal data and capture the global correlation between time series, thereby providing rich feature information for subsequent cancer detection. It can also extract low-rank factors in the data, effectively remove noise and redundant information in the signal, and improve the representation ability of the features. Compared with traditional feature extraction methods, the present invention can better maintain time series information, improve the performance of the model, and enhance its generalization ability in practical applications.

[0040] (3) The deep learning framework of the present invention has strong scalability and can process a large amount of electrical signal data and multi-dimensional physiological information, and is suitable for large-scale data analysis in clinical practice. In particular, it can achieve rapid and accurate cancer detection in early cancer screening and health monitoring, providing effective support for clinical diagnosis.

[0041] (4) By combining the deep learning model with the bioelectrical impedance detection technology, the present invention can not only improve the diagnostic accuracy but also reduce human intervention compared with the traditional bioelectrical impedance detection method. It has strong technical innovation and application potential and is suitable for the early screening and detection of various cancers such as cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of a bioelectrical impedance cancer detection method based on deep learning provided in an embodiment of the present invention; Figure 2 A schematic diagram of a matrix decomposition module flow chart of a bioelectrical impedance cancer detection method based on deep learning provided in an embodiment of the present invention; Figure 3 A feature fusion schematic diagram of a bioelectrical impedance cancer detection method based on deep learning provided in an embodiment of the present invention; Figure 4 Schematic diagram of feature extraction of a bioelectrical impedance cancer detection method based on deep learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0045] In the present invention, unless otherwise clearly specified and limited, the terms "connection", "fixation", etc. should be understood in a broad sense. For example, "fixation" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0046] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the meaning of "and / or" appearing in the full text includes three parallel schemes. Taking "A and / or B" as an example, it includes scheme A, or scheme B, or a scheme that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0047] Reference Figure 1-Figure 4 In a first aspect, the present invention provides a bioelectrical impedance cancer detection method based on deep learning, the method comprising the following steps: S1: using an electrical impedance imaging system to obtain electrical signal data of benign and malignant cancer tissue samples, and combining them with physiological data; the electrical signal data includes electrical parameters, electrical impedance data, and frequency response data; The present invention uses an electrical impedance imaging system to obtain electrical signal data of cancer tissues. By measuring and analyzing the electrical signal data of cancer tissues, information about tissue electrical parameters, electrical impedance data, and frequency response data can be obtained. Combining individual physiological data (such as age, gender, body mass index, etc.) helps eliminate interference caused by individual differences, can fully reflect the electrical characteristics of tissues, and help the model better distinguish between normal and cancerous tissues.

[0048] S2: Use the matrix decomposition module to extract global features of the electrical signal. The matrix decomposition module can effectively decompose the global time series features of the electrical signal data, capture the long-term dependencies and global trends in the electrical signal, and thus better understand the overall changes of the electrical signal. By extracting global features, the model can maintain good performance on different data sets, avoid overfitting problems, and enhance its generalization ability in practical applications.

[0049] S3: taking the result of the global feature extraction together with the original electrical signal data as new input, and further extracting electrical signal features by using data dependency; In the present invention, by combining global features with original signal data, the model can integrate global features and local details to improve the understanding and analysis capabilities of electrical signals. This fusion method can better capture cancer-related features and further extract electrical signal features using data dependencies, allowing the model to adaptively select the most important features and further improve classification performance.

[0050] S4: Integrate the extracted feature maps into the final classification results.

[0051] In this application, global features provide a macroscopic understanding of the signal, while the original signal data is used as local information and the extracted original signal features are used as local features to focus on the microscopic details of the signal; the combination of the two can make up for the limitations of a single feature and allow the model to not only consider long-term patterns when classifying, but also to keenly capture small, local abnormal changes, thereby better identifying cancer signals.

[0052] The present invention uses an electrical impedance imaging system to obtain electrical signal data of cancer, performs global feature extraction on the electrical signal through a matrix decomposition module, uses the result of global feature extraction together with the original electrical signal data as a new input, further extracts electrical signal features using data dependencies, and integrates the extracted feature mapping into the final classification result. The present invention has the advantages of high-precision data acquisition, effective signal processing, and fusion of global features and local features, and is expected to play an important role in the early detection and diagnosis of cancer.

[0053] As an embodiment of the present application, step S1 specifically includes the following steps: S11: Associating different samples with physiological data, wherein the physiological data includes age, gender, personal bad habits and body mass index; S12: The electrical impedance amplitude and phase at multiple frequencies are measured and a frequency response curve is drawn to reflect the tissue's response to currents of different frequencies.

[0054] Specifically, the present application can comprehensively reflect the electrical characteristics of tissues by combining individual physiological data (such as age, gender, body mass index, etc.), helping the model to better distinguish between normal and cancerous tissues. By correlating electrical signal data with physiological factors, it helps to remove the effects of factors such as smoking and obesity on electrical impedance signals, allowing the model to focus more on the characteristics of cancer and improve detection sensitivity.

[0055] like Figure 2 As shown, as an embodiment of the present application, the step S2 specifically includes: S21: First, the input electrical signal data is processed into time series data by a 1X1 convolutional layer; S22: decomposing the electrical signal processed by the 1X1 convolutional layer into low-rank factors and basis sequences; S23: Use the temporal convolutional network to extract time-dependent features from the basis sequence, and combine it with the low-rank factor for global feature fusion to achieve global feature extraction.

[0056] As an embodiment of the present application, the calculation formula for decomposing the electrical signal processed by the 1×1 convolution layer into low-rank factors and basis sequences in step S22 is as follows:

[0057]

[0058] in, is the input electrical signal data, is a low-rank factor used to capture the global correlation between signals, It is a base sequence that reflects the temporal pattern of the signal, helps compress the data dimension, and extracts more representative global features. is the electrical signal dimension, is the number of time steps.

[0059] Specifically, the present application can effectively decompose the global timing characteristics of electrical signal data through the matrix decomposition module, capture the long-term dependencies and global trends in the signal, and thus better understand the overall changes of the signal.

[0060] As an embodiment of the present application, step S23 specifically includes: S231: Base sequence The input is passed to the temporal convolutional network, and the long-term dependencies in the time dimension are captured through the extended convolution and residual connection of the temporal convolutional network. The convolution operation formula of each layer is as follows:

[0061] in, is the number of layers, is the dilation factor of the current layer, is the weight of the convolution kernel, is the output of the previous layer; The residual connection operation formula is as follows:

[0062] in, yes No. The output of the layer, is the input of the previous layer; specifically, the residual connection operation integrates the output of the convolution and the input of the previous layer. Through the residual connection plus the input of the previous layer, the model can gradually build a deeper understanding of the time-dependent characteristics of electrical signals between different layers, thereby better achieving the goals of global feature extraction and cancer detection.

[0063] S232: Low rank factor and the basis sequence processed by the temporal convolutional network Fusion is performed and splicing is performed on the feature dimension to form a complete global feature , the calculation formula is as follows:

[0064]

[0065] in, is the concatenated global feature, and its dimension is , represents the original electrical signal dimension, Represents the dimension of global features, represents the number of time steps, including the low-rank factor and base sequence The fusion process captures the channel and time correlation of electrical signals.

[0066] Specifically, the matrix decomposition module can extract low-rank factors in the data, effectively remove noise and redundant information in the signal, and improve the representation ability of features. By extracting global features, the model can maintain good performance on different data sets, avoid overfitting problems, and enhance its generalization ability in practical applications.

[0067] As an embodiment of the present application, step S3 specifically includes: S31: splicing and fusing the result of global feature extraction with the original electrical signal data; S32: adding position codes to the spliced ​​and fused data obtained in step S31 to obtain final input data; S33: Input the input data obtained in step S32 into a model built based on a deep learning framework, which serves to further extract global features.

[0068] like Figure 3 As shown, as an embodiment of the present application, the calculation formula for splicing and fusing the global features with the original electrical signal data in step S31 is as follows:

[0069] in, is the newly entered data, is the original electrical signal data, It is a global feature; Global Features And the original electrical signal data Splicing is performed on the channel dimension or time dimension, and the new input data after splicing on the time dimension , the dimension is .

[0070] Specifically, by combining global features with raw signal data, the model can integrate global features and local details to improve the understanding and analysis of electrical signals. This fusion method can better capture cancer-related features.

[0071] As an embodiment of the present application, step S32 specifically includes: S321: Convert the input data into a suitable sequence format, concatenate according to the time step, and the number of time steps No change, feature dimension increased, new input data is ,in is the new feature dimension, indicating the length of the feature vector at each time step; S322: Use an absolute position encoder to add position information, and generate position information for each time step through a function. The position code is calculated by sine and cosine functions, and the calculation formula is as follows:

[0072]

[0073] in, is the positional encoding, is the time step, is the index of the feature dimension, is the feature dimension size, and the position encoding encodes each time step through sine and cosine functions; Encode the position Add to the new input data after fusion It is used to integrate the time location information into the features to obtain the final input data , and its calculation formula is as follows:

[0074] .

[0075] Specifically, the global features and the original electrical signal features are integrated into a unified input format to provide support for subsequent model processing. After feature concatenation, the local characteristics of the original electrical signal are retained, while the global correlation information in the time series is enhanced; the time position information is incorporated into the input features, so that the model can identify the order of different time steps in the input sequence; the periodic characteristics of sine and cosine enable the model to effectively capture the long-range dependencies between time steps.

[0076] like Figure 4 As shown, as an embodiment of the present application, the step S33 specifically includes: S331: Processed input data When input into the model, the input data is first transformed into Converted into query, key and value, the calculation formula is as follows:

[0077]

[0078] in, is the learned weight matrix, are matrices of queries, keys, and values, respectively, is the dimension of the key, As the input of the self-attention mechanism, it is used to calculate the dependencies between each time step; S332: Calculate the similarity between the query and the key, and use this similarity to weight the value. The calculation formula is as follows:

[0079]

[0080] in, is the dimension of the key, used for scaling; Use the function to convert the attention score into a weight, and the calculation formula is as follows:

[0081] in, Representation query and key The attention weights between them and the final weighted sum value; S333: Divide the query, key, and value into multiple heads for calculation, so as to capture the feature information in different subspaces in parallel. The calculation method of each head is the same as above, and finally the outputs of all heads are concatenated. The calculation formula is as follows:

[0082]

[0083] in, is the number of heads, and the output of each head is , is the output linear transformation matrix, is the output; S334: A neural network consisting of two layers of linear transformation is used to further extract features, and each layer of linear transformation is followed by Activation function, the calculation formula of the first layer linear transformation is as follows:

[0084] in, is the weight matrix of the first layer linear transformation, is the bias term; The calculation formula of the second-layer linear transformation is as follows:

[0085] in, The weight matrix of the second layer linear transformation, is the bias term, is the output of the neural network; S335: Outputting the neural network Perform residual connection and layer normalization to get the final output .

[0086] Specifically, by calculating the similarity between the query and the key, the model can understand the dependencies between different time steps. This mechanism helps the model focus on other time steps related to the current time step, thereby improving the model's ability to learn long-range dependencies; through the multi-head mechanism, the model can understand the data from multiple perspectives, integrate the information of different attention heads, and improve the expressiveness and accuracy of the final output features. Through two layers of linear transformation (with Combined with the activation function), the neural network can further extract and process the features after the attention mechanism. This process helps the model to nonlinearly map complex electrical signal features, making the final representation richer and deeper.

[0087] As an embodiment of the present application, step S4 specifically includes: After processing, the final output is obtained It can be used for downstream tasks. The output formula of the classification task is as follows:

[0088] in, is the output, The layer maps the output to a probability space for classification tasks. The calculation formula of the layer is as follows:

[0089] in, Represents the model for category The output value of represents the total number of categories, In Represents a given input , the model predicts the category probability; pass Each probability value output by the layer represents the possibility that the input electrical signal belongs to each category; ultimately, the model will select the category with the highest probability as the classification result.

[0090] Specifically, through the processing of multiple encoder layers, the model can learn more abstract and profound feature representations from the input data and capture the complex relationships in the data. These deep features will directly affect the results of the classification task. Ultimately, the output of the classification task is determined by The probability distribution generated by the layer is determined, so that the input electrical signal can be effectively classified into different categories. Finally, the output of the classification task is determined by The probability distribution generated by the layer is determined, so that the input signal can be effectively classified into different categories.

[0091] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A bioelectrical impedance cancer detection method based on deep learning, characterized in that: The method comprises the following steps: S1: using an electrical impedance imaging system to obtain electrical signal data of benign and malignant cancer tissue samples, and combining them with physiological data; the electrical signal data includes electrical parameters, electrical impedance data, and frequency response data; S2: extracting global features of the electrical signal using a matrix decomposition module; S3: taking the result of the global feature extraction together with the original electrical signal data as new input, and further extracting electrical signal features by using data dependency; S4: Integrate the extracted feature maps into the final classification results.

2. A bioelectrical impedance cancer detection method based on deep learning as claimed in claim 1, characterized in that: The step S1 specifically includes the following steps: S11: Associating different samples with physiological data, wherein the physiological data includes age, gender, personal bad habits and body mass index; S12: The electrical impedance amplitude and phase at multiple frequencies are measured and a frequency response curve is drawn to reflect the tissue's response to currents of different frequencies.

3. A bioelectrical impedance cancer detection method based on deep learning as claimed in claim 1, characterized in that: The step S2 specifically includes: S21: First, the input electrical signal data is processed into time series data by a 1X1 convolutional layer; S22: decomposing the electrical signal processed by the 1X1 convolutional layer into low-rank factors and basis sequences; S23: Use the temporal convolutional network to extract time-dependent features from the basis sequence, and combine it with the low-rank factor for global feature fusion to achieve global feature extraction.

4. A bioelectrical impedance cancer detection method based on deep learning as claimed in claim 3, characterized in that: The calculation formula for decomposing the electrical signal processed by the 1×1 convolution layer into low-rank factors and basis sequences in step S22 is as follows: in, is the input electrical signal data, is a low-rank factor used to capture the global correlation between signals, is the basis sequence, which is used to reflect the temporal pattern of the signal. is the electrical signal dimension, is the number of time steps.

5. A bioelectrical impedance cancer detection method based on deep learning as claimed in claim 4, characterized in that: The step S23 specifically includes: S231: Base sequence The input is passed to the temporal convolutional network, and the long-term dependencies in the time dimension are captured through the extended convolution and residual connection of the temporal convolutional network. The convolution operation formula of each layer is as follows: in, is the number of layers, is the dilation factor of the current layer, is the weight of the convolution kernel, is the output of the previous layer; The residual connection operation formula is as follows: in, yes No. The output of the layer, is the input of the previous layer; S232: Low rank factor and the basis sequence processed by the temporal convolutional network Fusion is performed and splicing is performed on the feature dimension to form a complete global feature , the calculation formula is as follows: in, is the concatenated global feature, and its dimension is , represents the original electrical signal dimension, Represents the dimension of global features, represents the number of time steps, including the low-rank factor and base sequence information.

6. The bioelectrical impedance cancer detection method based on deep learning as claimed in claim 1, characterized in that: The step S3 specifically includes: S31: splicing and fusing the result of global feature extraction with the original electrical signal data; S32: adding position codes to the spliced ​​and fused data obtained in step S31 to obtain final input data; S33: Input the final input data obtained in step S32 into a model built based on a deep learning framework to further extract global features.

7. A bioelectrical impedance cancer detection method based on deep learning as claimed in claim 6, characterized in that: The calculation formula for splicing and fusing the global features with the original electrical signal data in step S31 is as follows: in, is the newly entered data, is the original electrical signal data, It is a global feature; Global Features And the original electrical signal data Splicing is performed on the channel dimension or time dimension, and the new input data after splicing on the time dimension , the dimension is .

8. The bioelectrical impedance cancer detection method based on deep learning as claimed in claim 7, characterized in that: The step S32 specifically includes: S321: Convert the input data into a suitable sequence format, concatenate according to the time step, and the number of time steps No change, feature dimension increased, new input data is ,in is the new feature dimension, indicating the length of the feature vector at each time step; S322: Use an absolute position encoder to add position information, and generate position information for each time step through a function. The position code is calculated by sine and cosine functions, and the calculation formula is as follows: in, is the positional encoding, is the time step, is the index of the feature dimension, is the feature dimension size, and the position encoding encodes each time step through sine and cosine functions; Encode the position Add to the new input data after fusion It is used to integrate the time location information into the features to obtain the final input data , and its calculation formula is as follows: 。 9. A bioelectrical impedance cancer detection method based on deep learning as claimed in claim 8, characterized in that: The step S33 specifically includes: S331: Processed input data When input into the model, the input data is first transformed into Converted into query, key and value, the calculation formula is as follows: in, is the learned weight matrix, They are query, key and value, Yes Key The dimension of As the input of the self-attention mechanism, it is used to calculate the dependencies between each time step; S332: Calculate the similarity between the query and the key, and use this similarity to weight the value. The calculation formula is as follows: in, is the dimension of the key, used for scaling; Use the function to convert the attention score into a weight, and the calculation formula is as follows: in, Representation query and key The attention weights between them and the final weighted sum value; S333: Divide the query, key, and value into multiple heads for calculation, so as to capture the feature information in different subspaces in parallel. The calculation method of each head is the same as above, and finally the outputs of all heads are concatenated. The calculation formula is as follows: in, is the number of heads, and the output of each head is , is the output linear transformation matrix, is the output; S334: A neural network consisting of two layers of linear transformation is used to further extract features, and each layer of linear transformation is followed by Activation function, the calculation formula of the first layer linear transformation is as follows: in, is the weight matrix of the first layer linear transformation, is the bias term; The calculation formula of the second-layer linear transformation is as follows: in, The weight matrix of the second layer linear transformation, is the bias term, is the output of the neural network; S335: Output of the neural network Perform residual connection and layer normalization to get the final output .

10. The bioelectrical impedance cancer detection method based on deep learning as claimed in claim 9, characterized in that: The step S4 specifically includes: After processing, the final output is For downstream tasks, the classification task output formula is as follows: in, is the output, The layer maps the output to a probability space for classification tasks. The calculation formula of the layer is as follows: in, Represents the model for category The output value of represents the total number of categories, In Represents a given input , the model predicts the category probability; pass Each probability value output by the layer represents the possibility that the input electrical signal belongs to each category; ultimately, the model will select the category with the highest probability as the classification result.

Images