Intracellular iron ion concentration detection method and system based on deep learning

Abstract

The invention provides an intracellular iron ion concentration detection method and system based on deep learning, and relates to the technical field of iron ion detection.The method comprises the steps that pH time sequence data and absorbance time sequence data of a reaction system in an intracellular iron ion chromogenic reaction are obtained; obtaining the final absorbance of the sample after the chromogenic reaction; carrying out feature extraction on the pH time sequence data, absorbance time sequence data and final absorbance of the reaction system, and training the time sequence convolutional network and the full connection layer fusion model by adopting a feature extraction result to obtain a prediction model; and obtaining pH time sequence data, absorbance time sequence data and final absorbance of a to-be-detected sample in a chromogenic reaction, extracting features, and inputting the features into the prediction model to obtain the iron ion concentration. The model constructed by the method can learn subtle changes in the reaction process, time sequence dynamics information and static concentration associated information are taken into account, and the accuracy of concentration calculation is improved.

Description

Technical Field

[0001] This invention relates to the field of iron ion detection technology, and more specifically, to a method and system for detecting intracellular iron ion concentration based on deep learning. Background Technology

[0002] Ferroprelation, an iron-dependent form of cell death, is closely related to the development and progression of various diseases, including cancer treatment, metabolic diseases, neurodegenerative diseases, and cardiovascular diseases. This conclusion is supported by existing research, and the association between ferroptosis and disease has been further validated through the construction and application of databases such as FerrDb. Activating or inhibiting ferroptosis can intervene in disease progression. Therefore, the detection of iron ions in tissues and cells has significant practical implications for scientific research and clinical treatment.

[0003] Currently, common laboratory methods for determining iron concentration include fluorescent probe methods and colorimetric analysis. Fluorescent probe methods determine the iron content in a sample by measuring the change in fluorescence intensity emitted after iron ions bind to a specific probe. Colorimetric analysis uses various metal chelates, such as bispyridine, as chromogenic agents, and quantifies the iron content in the sample by detecting the colored substance formed by the reaction of the chromogenic agent with iron ions.

[0004] However, each of the above detection methods has its own drawbacks. Fluorescent probe methods are not suitable for large-scale sample testing in general research institutions due to their high detection costs and sophisticated equipment requirements. Furthermore, the high price of reagent kits makes them difficult to widely promote. Colorimetric methods face multiple challenges, including poor reagent stability, short reagent shelf life, numerous interfering factors, and low detection accuracy for low-concentration samples. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting intracellular iron ion concentration based on deep learning, in order to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows: In a first aspect, this application provides a method for detecting intracellular iron ion concentration based on deep learning, comprising: In the intracellular iron ion colorimetric reaction, the pH time-series data and absorbance time-series data of the reaction system were obtained. The reagents for the iron ion colorimetric reaction included a copper ion chelating agent. The cell lysate sample of the colorimetric reaction was obtained by incubation with acidic buffer and potassium permanganate. Obtain the final absorbance of the sample after the colorimetric reaction; Feature extraction was performed on the pH time-series data, absorbance time-series data, and final absorbance of the reaction system. The feature extraction results were used to train the temporal convolutional network and the fully connected layer fusion model to obtain the prediction model. The pH time series data, absorbance time series data and final absorbance of the sample to be tested for colorimetric reaction are obtained, the features are extracted and input into the prediction model to obtain the iron ion concentration.

[0006] Secondly, this application also provides a deep learning-based intracellular iron ion concentration detection system, comprising: The first acquisition module is used to acquire the pH time series data and absorbance time series data of the reaction system in the intracellular iron ion colorimetric reaction. The reagents for the iron ion colorimetric reaction include copper ion chelating agents. The cell lysate sample of the colorimetric reaction is obtained by incubation with acidic buffer and potassium permanganate. The second acquisition module is used to acquire the final absorbance of the sample after the colorimetric reaction; The training module is used to extract features from the pH time-series data, absorbance time-series data, and final absorbance of the reaction system. The feature extraction results are used to train the temporal convolutional network and the fully connected layer fusion model to obtain the prediction model. The prediction module is used to acquire pH time-series data, absorbance time-series data and final absorbance of the sample to be tested for colorimetric reaction, extract features and input them into the prediction model to obtain the iron ion concentration.

[0007] The beneficial effects of this invention are as follows: This method enhances the iron ion release rate using potassium permanganate and eliminates specific interferences using copper ion chelating agents. It combines multi-dimensional data with deep learning. Compared with traditional colorimetric methods that calculate single signals in segments, the model constructed in this application can learn subtle changes in the reaction process, taking into account both temporal dynamics information and static concentration correlation information, thereby improving the accuracy of concentration identification and achieving high detection accuracy with simple detection methods.

[0008] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. Attached Figure Description

[0009] Figure 1 This is a flowchart of a deep learning-based method for detecting intracellular iron ion concentration according to an embodiment of this application. Figure 2 This is a schematic diagram of a deep learning-based intracellular iron ion concentration detection device according to an embodiment of this application.

[0010] The diagram is labeled as follows: 800 - Deep learning-based intracellular iron concentration detection device; 801 - Processor; 802 - Memory; 803 - Multimedia component; 804 - I / O interface; 805 - Communication component. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0012] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0013] Example 1 See Figure 1 This embodiment provides a method for detecting intracellular iron ion concentration based on deep learning, including steps S100, S200, S300 and S400; S100. Obtain pH and absorbance time-series data of the reaction system in the intracellular iron ion colorimetric reaction; the reagents for the iron ion colorimetric reaction include a copper ion chelating agent; the cell lysate sample of the colorimetric reaction is obtained by incubation with acidic buffer and potassium permanganate. pH time-series data were collected using a pH electrode sensor at preset time intervals, and absorbance time-series data were collected using a spectrophotometer (wavelength 550nm) at preset time intervals. To remove data noise, this step preprocesses the data, including: The pH time series data and absorbance time series data were respectively Z-score normalized. The decomposition level is determined based on the time series length, and wavelet decomposition is performed. An adaptive threshold is calculated for each decomposition level based on Stein unbiased risk estimation. Noise was removed using an adaptive threshold, resulting in denoised pH and absorbance time-series data.

[0014] The intracellular iron concentration detection reagents used in this step include: acidic buffer solution, potassium permanganate, ferroazine reagent, copper ion chelating agent, and reducing agent; The acidic buffer solution and potassium permanganate are used for iron ion extraction; the ferriazine reagent, copper ion chelating agent, and reducing agent are used for iron ion detection (colorimetric reaction).

[0015] Acidic buffer solutions include at least one of hydrochloric acid solution, oxalic acid solution, and citric acid solution, with a concentration of 0.1-2 mol / L; The concentration of potassium permanganate is 0.1-0.3 mol / L; The volume ratio of acidic buffer to potassium permanganate is 1:(0.5-2). The resulting acidic potassium permanganate solution is used to promote the separation of iron ions from proteins, achieving a more thorough extraction of iron ions from tissue cells. Proteins not treated with acidic potassium permanganate solution cannot completely release iron, which will negatively affect the accuracy of the results.

[0016] The concentration of ferroazine reagent is 8-20 mmol / L. The complex formed by ferroazine and iron ions has a strong orange spectral characteristic in the visible light region. According to the basic principle of colorimetry, the concentration of iron ions is determined by quantitatively measuring the change in absorbance of the complex in the solution.

[0017] Optionally, the copper ion chelating agent includes at least one of calcium magnesium reagent, ethylenediaminetetraacetic acid (EDTA), and neocopper reagent. In colorimetric detection of low-concentration samples, the presence of copper ions can lead to overestimation of the detection results, thus affecting the accuracy of the final results. This invention incorporates a copper ion chelating agent into the detection reagent, effectively eliminating the interference of copper ions on the detection system and ensuring the authenticity and accuracy of the detection results. The preferred concentration of the copper ion chelating agent is 6-20 mmol / L. Optionally, the reducing agent includes at least one selected from ascorbic acid, dithiothreitol (DTT), and mercaptoethanol. The concentration of the reducing agent is preferably 0.5-2 mol / L. This step also provides a deep learning-based method for detecting intracellular iron ion concentration, including: The sample was treated with lysis buffer, and then incubated for the first time in acidic buffer and potassium permanganate. After the first incubation, the sample was cooled to room temperature, and ferriazine reagent, copper ion chelating agent and reducing agent were added for a second incubation. The second incubation is the iron ion color reaction. Measure the final absorbance of the sample after the second incubation.

[0018] Optionally, the lysis solution includes at least one of potassium hydroxide and sodium hydroxide, with a concentration of 1-50 mmol / L. In some embodiments, the concentration may be 2 mmol / L, 10 mmol / L, 30 mmol / L, 35 mmol / L, 40 mmol / L, or 45 mmol / L.

[0019] Optionally, the temperature of the first incubation is 58-63℃, with 60℃ being optimal, which can improve the iron ion extraction efficiency. The incubation time is 0.8-1.2h, preferably 1h.

[0020] Optionally, the second incubation is carried out at room temperature for 0.4-0.7 hours, preferably 0.4 hours.

[0021] S200, Obtain the final absorbance of the sample after the colorimetric reaction; The concentration of iron ions can usually be determined by comparing the final absorbance with the standard absorbance curve. However, this traditional method may be subject to errors caused by impurities. Moreover, for the detection of low concentration iron ions, small concentration differences may not cause changes in absorbance. Therefore, the results obtained by comparing with the standard curve are not very accurate. This application proposes to achieve precise quantification by combining multi-dimensional data with deep learning. Since the pH value of the reaction system changes over time after the addition of potassium permanganate, it affects the binding of ferroazine and iron ions. Therefore, there is a certain correlation between the iron ion concentration, pH change and absorbance change. This method establishes a mapping relationship with the true iron ion concentration by mining the subtle changes in the reaction process and combining them with the final absorbance, thereby improving the accuracy of concentration detection. It is especially suitable for the detection of low concentration samples.

[0022] S300: Extract features from the pH time-series data, absorbance time-series data, and final absorbance of the reaction system; train the temporal convolutional network and the fully connected layer fusion model to obtain the prediction model. For low-concentration samples, the numerical differences in their time-series data may be small, but the data structures are different. Therefore, this application extracts the topological structure of the data (such as inflection point positions and curve smoothness) to obtain implicit structural information. The structural information complements the numerical information, thereby effectively improving the discriminative power of low-concentration features. The specific steps are as follows: S310. Convert the pH time series data and absorbance time series data into point cloud structures respectively; S320. Use a preset window to slide and capture point cloud structure data, and extract the topological structure features of pH and absorbance respectively. As an example, the pH data and absorbance data were collected at the same frequency, with a window size of 5 time points and a sliding step size of 1. Several small sequences were extracted, and the topological features of each small sequence, such as peak and valley positions and height differences, were extracted. S330. The topological features of pH are concatenated with pH time-series data, and the topological features of absorbance are concatenated with absorbance time-series data to obtain structure enhancement features. The original time-series data is responsible for learning the dynamic changes of the time series, while the topological features are responsible for providing structural information, thus solving the problem that the time-series numerical differences of low concentrations are small but the structures are different.

[0023] To identify the contribution of different reaction stages to iron ion detection, this method also includes: S340. The topological feature sequences of pH and absorbance are encoded by a bidirectional long short-term memory network to obtain the encoded features of each window. The static topological features are transformed into dynamic features with time-dependent characteristics by a bidirectional long short-term memory network to strengthen local time-dependent characteristics. S350. Calculate the attention weight matrix for each encoded feature between pH and absorbance using a cross-modal attention mechanism; For example, the attention weight matrix is ​​W[i,j], which represents the association weight between pH window i and absorbance window j. The larger the value, the stronger the joint effect of the two in concentration recognition.

[0024] S360. Divide the total reaction time into multiple time periods according to the reaction stage. Calculate the total weight of each time period based on the attention weight matrix of each encoded feature. Weight the features of pH data and absorbance data in different time periods. Divide the N windows into 3-5 time periods according to the reaction stage (such as early stage, middle stage, and late stage). The total weight of each time period can reflect the contribution of each time period to the final detection result. For time periods with a large contribution, the data of that stage can be weighted, and the data collection frequency of that stage can be shortened.

[0025] Traditional temporal convolutional networks, with their single convolutional kernel size, can only capture temporal dependencies within a fixed range. However, the effective information in low-concentration temporal data may be distributed across different time scales, such as small fluctuations in the early stages of a reaction and a long-term upward trend in the later stages. To comprehensively capture this effective information, this application uses 3-4 convolutional kernels of different sizes to process the same temporal input in parallel, extracting feature information at different time scales, as detailed below: A temporal convolutional network with multi-scale parallel convolutional branches is constructed. Each convolutional branch adopts causal convolution. pH time-series data and absorbance time-series data are input into the temporal convolutional network to obtain global feature vectors at different scales. In this step, three convolutional branches are set with kernel sizes of 2, 5 and 8, respectively, to gradually expand the temporal receptive field and capture a more global trend. Calculate the output weights of different convolution branches based on the total weight distribution at different time points in the reaction process; When the weights differ significantly across different time periods in the reaction process, focus on short-term local data features by increasing the output weights of smaller convolutional kernels. Conversely, when the weight differences are smaller across different time periods, increase the output weights of larger convolutional kernels to focus more on long-term global trends.

[0026] The global feature vectors of each branch are dynamically weighted and fused based on the output weights to obtain scale-fused features. Channels assign dynamic weights to output features at different scales, avoiding feature redundancy caused by simple concatenation.

[0027] In this step, the output samples of the training set are the iron ion concentrations obtained by detecting the samples using the fluorescent probe method, so that the output of the trained model can achieve detection accuracy similar to that of the fluorescent probe method.

[0028] Differences in cell sample types (such as protein composition and endogenous interfering substance content) may lead to a decrease in the model's generalization ability. Therefore, it is necessary to minimize the differences caused by interference from different cell types during feature extraction. This application trains the feature extractor through an adversarial network to make the feature distribution of different cells more consistent, thereby reducing prediction bias caused by different sample sources, as detailed below: Obtain cell type information, construct a cell classifier, and construct a loss function for the cell classifier with the goal of minimizing cell classification error; A feature extraction network was used to extract features from the pH time-series data, absorbance time-series data, and final absorbance of the reaction system. The extracted features were then input into a cell classifier, and the classification loss function was calculated. When the classification loss of the cell classifier decreases, the feature extraction network is updated in reverse with a preset negative coefficient gradient, which reduces the encoding of cell-related features by the feature extractor, and ultimately makes the distribution of shared features of different cells overlap.

[0029] S400: Acquire pH time-series data, absorbance time-series data, and final absorbance of the sample to be tested for colorimetric reaction, extract features, and input them into the prediction model to obtain the iron ion concentration.

[0030] The method of this application is used to detect the concentration of iron ions in cells. Specific embodiments are as follows: Cells were lysed on a shaker with 100 μL of lysis buffer. After lysis, the cells were centrifuged, and 100 μL of supernatant was collected. The lysis buffer was 20 mmol / L sodium hydroxide. The supernatant was divided into three portions, and two of them were diluted and labeled A, B, and C. The following operations were then performed on each portion: Add 50 μL of acidic buffer and 50 μL of potassium permanganate (0.30 mol / L) to the supernatant, mix well, and incubate at 60 °C for 1 h; the acidic buffer is a 1 mol / L hydrochloric acid solution. After the first incubation, the sample was cooled to room temperature, and 30 μL of a mixed reagent was added. The mixed reagent contained 10 mmol / L ferroazine, 8 mmol / L copper ion chelating agent (new copper ion reagent), and 1 mmol / L reducing agent (ascorbic acid). A second incubation was then performed for 30 min. The pH and absorbance data were measured during the second incubation. The final absorbance of the sample was measured after the second incubation.

[0031] The prediction model was constructed by inputting pH time series data, absorbance time series data, and final absorbance to obtain the iron ion concentration.

[0032] The results of the iron ion concentration determination for samples A, B, and C are shown in Table 1: Table 1

[0033] It can be observed that the prediction model constructed in this application has a small error with the fluorescent probe, and its recognition precision and accuracy are higher than those of the traditional colorimetric method, whose lowest detection limit is generally only 1-5 μmol / L.

[0034] Example 2 This embodiment provides a deep learning-based intracellular iron ion concentration detection system, comprising: The first acquisition module is used to acquire the pH time series data and absorbance time series data of the reaction system in the intracellular iron ion colorimetric reaction. The reagents for the iron ion colorimetric reaction include copper ion chelating agents. The cell lysate sample of the colorimetric reaction is obtained by incubation with acidic buffer and potassium permanganate. The second acquisition module is used to acquire the final absorbance of the sample after the colorimetric reaction; The training module is used to extract features from the pH time-series data, absorbance time-series data, and final absorbance of the reaction system. The feature extraction results are used to train the temporal convolutional network and the fully connected layer fusion model to obtain the prediction model. The prediction module is used to acquire pH time-series data, absorbance time-series data and final absorbance of the sample to be tested for colorimetric reaction, extract features and input them into the prediction model to obtain the iron ion concentration.

[0035] As an optional implementation, the training module includes: The conversion unit is used to convert pH time-series data and absorbance time-series data into point cloud structures, respectively. The extraction unit is used to slide and capture point cloud structure data with a preset window, and extract the topological structure features of pH and absorbance respectively. The splicing unit is used to splice the topological features of pH with pH time-series data, and to splice the topological features of absorbance with absorbance time-series data to obtain structural enhancement features.

[0036] As an optional implementation, the system further includes: The encoding module is used to encode the topological feature sequences of pH and absorbance respectively through a bidirectional long short-term memory network to obtain the encoded features of each window; The first computation module is used to compute the attention weight matrix of each encoded feature between pH and absorbance through a cross-modal attention mechanism; The second calculation module is used to divide the total reaction time into multiple time periods according to the reaction stage, calculate the total weight of each time period based on the attention weight matrix of each encoded feature, and weight the features of pH data and absorbance data in different time periods.

[0037] As an optional implementation, the training module further includes: The construction unit is used to construct a temporal convolutional network with multi-scale parallel convolutional branches. Each convolutional branch adopts causal convolution and inputs pH time-series data and absorbance time-series data into the temporal convolutional network to obtain global feature vectors at different scales. The computational unit is used to calculate the output weights of different convolutional branches based on the total weight distribution at different time points in the reaction process. The fusion unit is used to dynamically weight and fuse the global feature vectors of each branch according to the output weights to obtain scale-fused features.

[0038] Example 3 Corresponding to the above method embodiments, this embodiment also provides a deep learning-based intracellular iron ion concentration detection device. The deep learning-based intracellular iron ion concentration detection device described below can be referred to in conjunction with the deep learning-based intracellular iron ion concentration detection method described above.

[0039] Figure 2 This is a block diagram illustrating a deep learning-based intracellular iron ion concentration detection device 800 according to an exemplary embodiment. Figure 2As shown, the deep learning-based intracellular iron concentration detection device 800 includes a processor 801 and a memory 802. The deep learning-based intracellular iron concentration detection device 800 may also include one or more of the following: a multimedia component 803, an input / output (I / O) interface 804, and a communication component 805. The processor 801 controls the overall operation of the deep learning-based intracellular iron concentration detection device 800 to complete all or part of the steps in the aforementioned deep learning-based intracellular iron concentration detection method. The memory 802 stores various types of data to support the operation of the deep learning-based intracellular iron concentration detection device 800. This data may include, for example, commands for any application or method operating on the deep learning-based intracellular iron concentration detection device 800, and application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0040] Multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals.

[0041] The received audio signal can be further stored in memory 802 or transmitted via communication component 805. The audio component also includes at least one speaker for outputting the audio signal. I / O interface 804 provides an interface between processor 801 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the deep learning-based intracellular iron ion concentration detection device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof, is possible; therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0042] Example 4 Corresponding to the above embodiment of the deep learning-based intracellular iron ion concentration detection method, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the deep learning-based intracellular iron ion concentration detection method described above.

[0043] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described embodiment of the deep learning-based intracellular iron ion concentration detection method.

[0044] Specifically, the readable storage medium can be a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, or any other readable storage medium capable of storing program code.

[0045] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0046] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting intracellular iron ion concentration based on deep learning, characterized in that, include: In the intracellular iron ion colorimetric reaction, the pH time-series data and absorbance time-series data of the reaction system were obtained. The reagents for the iron ion colorimetric reaction included a copper ion chelating agent. The cell lysate sample of the colorimetric reaction was obtained by incubation with acidic buffer and potassium permanganate. Obtain the final absorbance of the sample after the colorimetric reaction; Feature extraction was performed on the pH time-series data, absorbance time-series data, and final absorbance of the reaction system. The feature extraction results were used to train the temporal convolutional network and the fully connected layer fusion model to obtain the prediction model. The pH time series data, absorbance time series data and final absorbance of the sample to be tested for colorimetric reaction are obtained, the features are extracted and input into the prediction model to obtain the iron ion concentration.

2. The method for detecting intracellular iron ion concentration based on deep learning according to claim 1, characterized in that, The feature extraction of the reaction system's pH time-series data, absorbance time-series data, and final absorbance includes: pH time series data and absorbance time series data were converted into point cloud structures respectively; Point cloud structure data is captured by sliding a preset window, and the topological features of pH and absorbance are extracted respectively. The topological features of pH and pH time-series data are concatenated, and the topological features of absorbance and absorbance time-series data are concatenated to obtain the structure enhancement features.

3. The method for detecting intracellular iron ion concentration based on deep learning according to claim 2, characterized in that, The method further includes: The topological feature sequences of pH and absorbance are encoded using a bidirectional long short-term memory network to obtain the encoded features of each window. The attention weight matrix for each encoded feature between pH and absorbance is calculated using a cross-modal attention mechanism; The total reaction time is divided into multiple time periods according to the reaction stage. Based on the attention weight matrix of each encoded feature, the total weight of each time period is calculated, and the features of pH data and absorbance data in different time periods are weighted.

4. The method for detecting intracellular iron ion concentration based on deep learning according to claim 3, characterized in that, The feature extraction of the reaction system's pH time-series data, absorbance time-series data, and final absorbance includes: A temporal convolutional network with multi-scale parallel convolutional branches is constructed. Each convolutional branch adopts causal convolution. pH time-series data and absorbance time-series data are input into the temporal convolutional network to obtain global feature vectors at different scales. Calculate the output weights of different convolution branches based on the total weight distribution at different time points in the reaction process; The global feature vectors of each branch are dynamically weighted and fused based on the output weights to obtain scale-fused features.

5. The method for detecting intracellular iron ion concentration based on deep learning according to claim 1, characterized in that, The method further includes: Obtain cell type information, construct a cell classifier, and construct a loss function for the cell classifier with the goal of minimizing cell classification error; A feature extraction network was used to extract features from the pH time-series data, absorbance time-series data, and final absorbance of the reaction system. The extracted features were then input into a cell classifier, and the classification loss function was calculated. When the classification loss of the cell classifier decreases, the feature extraction network is updated in reverse with a preset negative coefficient gradient.

6. The method for detecting intracellular iron ion concentration based on deep learning according to claim 1, characterized in that, The method further includes: The pH time series data and absorbance time series data were respectively Z-score normalized. The decomposition level is determined based on the time series length, and wavelet decomposition is performed. An adaptive threshold is calculated for each decomposition level based on Stein unbiased risk estimation. Noise was removed using an adaptive threshold, resulting in denoised pH and absorbance time-series data.

7. A deep learning-based intracellular iron ion concentration detection system, characterized in that, include: The first acquisition module is used to acquire the pH time series data and absorbance time series data of the reaction system in the intracellular iron ion colorimetric reaction. The reagents for the iron ion colorimetric reaction include copper ion chelating agents. The cell lysate sample of the colorimetric reaction is obtained by incubation with acidic buffer and potassium permanganate. The second acquisition module is used to acquire the final absorbance of the sample after the colorimetric reaction; The training module is used to extract features from the pH time-series data, absorbance time-series data, and final absorbance of the reaction system. The feature extraction results are used to train the temporal convolutional network and the fully connected layer fusion model to obtain the prediction model. The prediction module is used to acquire pH time-series data, absorbance time-series data and final absorbance of the sample to be tested for colorimetric reaction, extract features and input them into the prediction model to obtain the iron ion concentration.

8. The intracellular iron ion concentration detection system based on deep learning according to claim 7, characterized in that, The training module includes: The conversion unit is used to convert pH time-series data and absorbance time-series data into point cloud structures, respectively. The extraction unit is used to slide and capture point cloud structure data with a preset window, and extract the topological structure features of pH and absorbance respectively. The splicing unit is used to splice the topological features of pH with pH time-series data, and to splice the topological features of absorbance with absorbance time-series data to obtain structural enhancement features.

9. The intracellular iron ion concentration detection system based on deep learning according to claim 8, characterized in that, The system also includes: The encoding module is used to encode the topological feature sequences of pH and absorbance respectively through a bidirectional long short-term memory network to obtain the encoded features of each window; The first computation module is used to compute the attention weight matrix of each encoded feature between pH and absorbance through a cross-modal attention mechanism; The second calculation module is used to divide the total reaction time into multiple time periods according to the reaction stage, calculate the total weight of each time period based on the attention weight matrix of each encoded feature, and weight the features of pH data and absorbance data in different time periods.

10. The method for detecting intracellular iron ion concentration based on deep learning according to claim 9, characterized in that, The training module also includes: The construction unit is used to construct a temporal convolutional network with multi-scale parallel convolutional branches. Each convolutional branch adopts causal convolution and inputs pH time-series data and absorbance time-series data into the temporal convolutional network to obtain global feature vectors at different scales. The computational unit is used to calculate the output weights of different convolutional branches based on the total weight distribution at different time points in the reaction process. The fusion unit is used to dynamically weight and fuse the global feature vectors of each branch according to the output weights to obtain scale-fused features.

Images