A multi-modal molecular imaging method and device based on dynamic iterative calibration of DESI mass spectrometry

Abstract

This application provides a multimodal molecular imaging method and apparatus based on DESI mass spectrometry dynamic iterative calibration, which can balance high-throughput data acquisition and high-precision molecular imaging. The method includes: acquiring optical imaging data of a sample to be tested; selecting multiple calibration points from the sample to be tested based on the optical imaging data to obtain a sparse calibration point set; acquiring mass spectrometry imaging data of each calibration point in the sparse calibration point set; establishing a mapping model between the mass spectrometry imaging data and the corresponding optical imaging data using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as a reference true value; and calibrating the optical imaging data of the sample to be tested based on the mapping model to generate a molecular image of the sample to be tested.

Description

Technical Field

[0001] This application relates to the field of molecular imaging technology, and in particular to a multimodal molecular imaging method and apparatus based on DESI mass spectrometry dynamic iterative calibration. Background Technology

[0002] Hyperspectral imaging and Raman spectroscopy, with their rapid scanning capabilities, can efficiently acquire spatial distribution information of chemical components in samples and are widely used in large-scale sample detection scenarios. However, due to limitations in their technical principles, they have significant limitations in absolute quantitative analysis and precise identification of specific molecules, making it difficult to meet the application requirements for high accuracy of molecular information. Mass spectrometry provides precise molecular weight information and high spatial resolution, enabling accurate molecular identification and quantification, and is regarded as the "gold standard" in molecular imaging. However, this technology has a low data acquisition throughput, and completing a large-area sample imaging often takes a long time, making it unsuitable for rapid, large-area imaging applications.

[0003] Therefore, there is an urgent need for a multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration, which can take into account both high-throughput data acquisition and high-precision molecular imaging. Summary of the Invention

[0004] The multimodal molecular imaging method and device based on DESI mass spectrometry dynamic iterative calibration provided in this application can take into account both high-throughput data acquisition and high-precision molecular imaging.

[0005] In a first aspect, this application provides a multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration, the method comprising: Acquire optical imaging data of the sample to be tested; Based on the optical imaging data, a sparse calibration point set is obtained by selecting multiple calibration points from the sample to be tested. Obtain mass spectrometry imaging data of each calibration point in the sparse calibration point set; Using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value, a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding point is established. Based on the mapping model, the optical imaging data of the sample to be tested is calibrated to generate a molecular image of the sample to be tested.

[0006] Optionally, the method further includes: Determine whether the molecular image of the sample to be tested meets the iteration termination condition; If the iteration termination condition is not met, the following steps are performed until the iteration termination condition is met: Points in the molecular image of the sample to be tested with prediction uncertainty higher than a set threshold are added as new calibration points and added to the current sparse calibration point set to obtain an expanded sparse calibration point set. Acquire the mass spectrometry imaging data of the newly added calibration points; The current mapping model is optimized based on the expanded sparse calibration point set to obtain the optimized mapping model. Based on the optimized mapping model, the optical imaging data of the sample to be tested is calibrated to generate an optimized molecular image of the sample to be tested.

[0007] Optionally, the method further includes: If the iteration termination condition is met, the optimized molecular image of the sample to be tested will be output as the molecular image of the sample to be tested. The current mapping model and the relevant data of the optimized molecular images of the test samples are added to the molecular imaging database as samples.

[0008] Optionally, before selecting multiple calibration points from the sample under test based on the optical imaging data to obtain a sparse calibration point set, the method further includes: Calculate the similarity between the optical imaging data of the sample to be tested and the optical imaging data of the samples in the molecular imaging database; The mapping model corresponding to the sample with the highest similarity is selected as the initial mapping model; Based on the initial mapping model, the optical imaging data of the sample to be tested is calibrated to generate an initial molecular image of the sample to be tested. If the initial molecular image meets the imaging requirements, then the initial molecular image is output as the molecular image of the sample to be tested; If the initial molecular image does not meet the imaging requirements, the initial mapping model is optimized.

[0009] Optionally, the method further includes: The mapping model is established using the following method: The optical imaging data of each calibration point of the sample under test is used to generate noise data for each calibration point by adding noise to the optical imaging data of each calibration point through the forward generation module. The noise data, location information, and cell clustering code of each calibration point are input into the inverse noise reduction module to train the neural network in the inverse noise reduction module, so that the neural network can generate mass spectrometry imaging data corresponding to each calibration point of the sample to be tested. The neural network is used as the mapping model.

[0010] Optionally, the iteration termination condition includes at least one of the following: The molecular images of the sample to be tested meet the imaging requirements; The number of calibration points has reached the first preset threshold. The contribution of newly added calibration points to the optimization of the mapping model is less than the second preset threshold; The similarity between the optimized molecular image of the test sample and the baseline molecular image reaches the third set threshold.

[0011] Secondly, this application provides a multimodal molecular imaging device based on DESI mass spectrometry dynamic iterative calibration, the device comprising: The acquisition module is used to acquire optical imaging data of the sample under test. The processing module is used to select multiple calibration points from the sample under test based on the optical imaging data to obtain a sparse calibration point set; The acquisition module is also used to acquire mass spectrometry imaging data of each calibration point in the sparse calibration point set; The mapping model module is used to establish a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding points, using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value. The generation module is used to calibrate the optical imaging data of the sample under test based on the mapping model, and generate a molecular image of the sample under test.

[0012] Optionally, the processing module is further configured to determine whether the molecular image of the sample to be tested satisfies the iteration termination condition; If the iteration termination condition is not met, the following steps are performed until the iteration termination condition is met: Points in the molecular image of the sample to be tested with prediction uncertainty higher than a set threshold are added as new calibration points and added to the current sparse calibration point set to obtain an expanded sparse calibration point set. The acquisition module is also used to acquire the mass spectrometry imaging data of the newly added calibration point; The mapping model module is also used to optimize the current mapping model based on the expanded sparse calibration point set to obtain an optimized mapping model. The generation module is also used to calibrate the optical imaging data of the sample to be tested based on the optimized mapping model, and generate an optimized molecular image of the sample to be tested.

[0013] Optionally, the processing module is further configured to output the optimized molecular image of the sample to be tested as the molecular image of the sample to be tested if the iteration termination condition is met. The current mapping model and the relevant data of the optimized molecular images of the test samples are added to the molecular imaging database as samples.

[0014] Optionally, before selecting multiple calibration points from the sample to be tested based on the optical imaging data to obtain a sparse calibration point set, the processing module is further used to calculate the similarity between the optical imaging data of the sample to be tested and the optical imaging data of the samples in the molecular imaging database. The mapping model corresponding to the sample with the highest similarity is selected as the initial mapping model; The generation module is also used to calibrate the optical imaging data of the sample to be tested based on the initial mapping model, and generate an initial molecular image of the sample to be tested. If the initial molecular image meets the imaging requirements, the processing module is further configured to output the initial molecular image as the molecular image of the sample to be tested. If the initial molecular image does not meet the imaging requirements, the mapping model module is further used to optimize the initial mapping model.

[0015] Optionally, the mapping model module is further configured to establish a mapping model using the following method: The optical imaging data of each calibration point of the sample under test is used to generate noise data for each calibration point by adding noise to the optical imaging data of each calibration point through the forward generation module. The noise data, location information, and cell clustering code of each calibration point are input into the inverse noise reduction module to train the neural network in the inverse noise reduction module, so that the neural network can generate mass spectrometry imaging data corresponding to each calibration point of the sample to be tested. The neural network is used as the mapping model.

[0016] Optionally, the iteration termination condition includes at least one of the following: The molecular images of the sample to be tested meet the imaging requirements; The number of calibration points has reached the first preset threshold. The contribution of newly added calibration points to the optimization of the mapping model is less than the second preset threshold; The similarity between the optimized molecular image of the test sample and the baseline molecular image reaches the third set threshold.

[0017] Thirdly, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration as described above.

[0018] Fourthly, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration as described above.

[0019] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration as described above.

[0020] The multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration provided in this application adopts a "sparse sampling + algorithm calibration" strategy. It only needs to collect a small amount of mass spectrometry imaging data of sparse points in the sample to obtain results with near-complete mass spectrometry imaging quality. While ensuring the accuracy of molecular recognition, it greatly reduces the analysis time and achieves the unification of high-throughput data acquisition and high-precision molecular imaging. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 System architecture diagram provided for embodiments of this application; Figure 2 A flowchart illustrating a multimodal molecular imaging method based on dynamic iterative calibration of DESI mass spectrometry, provided for an embodiment of this application; Figure 3 A schematic diagram illustrating the process of establishing a mapping model as provided in an embodiment of this application; Figure 4 A schematic diagram of the overall process of a multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration provided in this application embodiment; Figure 5 A schematic diagram of a multimodal molecular imaging method device based on DESI mass spectrometry dynamic iterative calibration provided in this application embodiment; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] All actions involving the acquisition of signal information or data in this application are carried out in accordance with the relevant data protection laws and policies of the country where the application is located, and with the authorization of the owner of the relevant device.

[0025] In the embodiments of this application, "multiple" refers to two or more. Terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0026] Figure 1 The system architecture diagram provided for the embodiments of this application is as follows: Figure 1 As shown, the system includes an optical system, a mass spectrometry system, an intelligent mobile station, and an AI model.

[0027] The optical system uses autofocus technology to achieve real-time focusing on the sample to be tested, acquire a complete optical image of the sample, and obtain the optical signal of the sample, i.e., optical imaging data.

[0028] The mass spectrometry system uses an intelligent mobile stage to precisely locate calibration sites and acquire mass spectrometry imaging data at those sites. For example, desorption electrospray ionization (DESI) mass spectrometry imaging technology can be employed. Electrospray solvent forms charged microdroplets, which are sprayed onto the surface of the sample to be tested, extracting and ionizing molecules on the sample surface. The generated ions are transmitted to the mass spectrometry system via an ion conduction atmospheric interface for analysis, generating a mass spectrometry signal that provides a chemical fingerprint of the sample at a specific point, i.e., precise molecular composition information.

[0029] The AI ​​model acquires optical imaging data from the optical system and mass spectrometry imaging data from the mass spectrometry system. Using the mass spectrometry imaging data as a reference, a mapping model between the optical and mass spectrometry imaging data at calibration points is constructed using machine learning algorithms. Based on this model, a molecular image of the sample is generated, and multiple calibration points are used to ensure the molecular image accuracy meets imaging requirements. Based on the output molecular image of the sample, the system generates a rapid pathology report and constructs a molecular imaging reference database for initial calibration of subsequent sample testing.

[0030] Figure 2 A flowchart illustrating a multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration is provided for embodiments of this application, as shown below. Figure 2 As shown, the method includes the following steps: Step 210: Obtain optical imaging data of the sample to be tested.

[0031] The optical imaging data of the sample to be tested is acquired. This optical imaging data can be either hyperspectral imaging data or Raman spectral imaging data. Specifically, a high-resolution chemical composition distribution map of the sample to be tested can be obtained using a hyperspectral camera, i.e., hyperspectral imaging data, or Raman spectral imaging data can be obtained by scanning the sample to be tested using a confocal Raman microscope.

[0032] Step 220: Based on optical imaging data, select multiple calibration points from the sample to be tested to obtain a sparse calibration point set.

[0033] Based on optical imaging data, multiple calibration points representing chemical characteristics can be selected on the surface of the sample under test through chemical spatial heterogeneity analysis. Alternatively, a clustering algorithm can be used to divide the image of the sample under test into several regions with similar chemical characteristics, and one calibration point can be randomly selected from each region. Alternatively, multiple calibration points can be randomly selected from different regions of the sample under test. These multiple calibration points constitute a sparse calibration point set, where the number of calibration points is far less than the total number of pixels in the optical imaging data.

[0034] Step 230: Obtain mass spectrometry imaging data of each calibration point in the sparse calibration point set.

[0035] Mass spectrometry imaging data of each calibration point in the sparse calibration point set is acquired. The mass spectrometry imaging data can be mass spectrometry imaging data acquired by DESI mass spectrometry imaging technology. Specifically, a high-precision intelligent moving stage is used to control the DESI ion source to move to each calibration point and acquire the full mass spectrometry information of the calibration point.

[0036] This application does not specify the method of acquiring mass spectrometry imaging data. For example, mass spectrometry imaging data can also be acquired through mass spectrometry probe ionization techniques that meet specific standards, such as laser ablation electrospray ionization (LAESI), direct analysis in real time (DART) and its derivatives, secondary electrospray ionization (SESI), matrix-assisted laser desorption / ionization (MALDI), laser sputter ionization (LSI) and its variants.

[0037] Step 240: Using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value, establish a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding points.

[0038] Specifically, using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value, a machine learning algorithm is used to establish a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding points. For example, the ion intensity of characteristic molecules in the tissue of the sample to be tested can be selected as the dependent variable, and the intensity of the hyperspectral characteristic spectrum or Raman characteristic peak of the corresponding point can be selected as the independent variable. A mapping model X→Y is constructed based on the noise reduction diffusion probability model.

[0039] Figure 3 A schematic diagram illustrating the process of establishing a mapping model provided in this application embodiment, as shown below. Figure 3 As shown, the process of establishing the mapping model includes a forward generation process and a reverse decoding process: Forward generation process: The optical imaging data of each calibration point of the sample to be tested is used to add noise through the forward generation module to generate noise data for each calibration point.

[0040] Taking hyperspectral imaging data as an example, the forward generation module, during the forward process, targets the hyperspectral region image. Using the noise-adding function Add gradually Secondary noise is generated, producing a noise map of the same size as the original image. From the initial state... go through step( The probability of () can be written as: in, It is a set of sparse calibration points acquired from known optical images. These are unknown latent variables in the forward process; in this case, the joint probability of the model becomes a conditional probability. Each Each variable is a Gaussian variable, representing the encoder at each step of the forward process. Fixed to a linear Gaussian transform, i.e. mean and The value of is a linear relationship. This means that each step adds a random Gaussian noise data point to the previous step, and as ... The increase, It gradually becomes Gaussian noise data.

[0041] Reverse decoding process: Input the noise data of each calibration point, the location information of each calibration point and the cell clustering code into the reverse denoising module, train the neural network in the reverse denoising module, so that the neural network can generate the mass spectrometry imaging data corresponding to each calibration point of the sample to be tested, and use the neural network as a mapping model.

[0042] The location information of each calibration point and the cell clustering code are generated through a conditional control module. Since the forward pass essentially adds noise to the training data, the conditional control module primarily influences intermediate variables to facilitate subsequent inverse noise reduction and mass spectrometry imaging data generation. These intermediate variables include location encoding information encoded using a sinusoidal method. and the clustering encoding of the corresponding cells in that region. Both of them are connected to the intermediate vector. Together, they serve as input to the inverse noise reduction module. Specifically, the location information of this region within the complete hyperspectral data is first obtained. These represent the horizontal and vertical channel coordinates, as well as the width and height, respectively, and their positional encoding information. It can be represented as: in , , which is the standard Sinusoidal positional encoding function, with the dimension set to 512.

[0043] It can be represented as: in, The pre-trained hyperspectral image representation model can effectively represent different tissues (e.g., tumors, non-tumors) as a vector of size 1024.

[0044] The inverse noise reduction module removes random Gaussian noise. Starting with location coding information and cell clustering coding information The image is gradually decoded to form the corresponding mass spectrometry image in this invention. Each moment of the decoding process... To Restore to The conditional probability corresponding to each step is expressed as Following the reverse process, the joint probability... Decompose into: in, It is a standard Gaussian distribution, and the conditional probability distribution is learned by fitting it using a neural network, from a random Gaussian noise. The corresponding mass spectrometry images are generated step by step. .

[0045] In addition, machine learning algorithms such as partial least squares regression, support vector regression, or neural network models can be used to establish mapping models, and this application does not make specific limitations on this.

[0046] Step 250: Based on the mapping model, calibrate the optical imaging data of the sample to be tested to generate a molecular image of the sample to be tested.

[0047] Based on the mapping model established in step 240, the optical imaging data of the sample to be tested is calibrated to generate a molecular image of the sample. Specifically, during the inference process using the mapping model, the weights of the mapping model are first fixed. Then, the region encoding of the optical imaging data of the sample to be tested and the cell clustering encoding result of that region, along with randomly generated noise, are input into the mapping model. The mapping model is then applied to the overall optical image to predict the characteristic molecular ion intensity of each pixel, thereby generating a molecular image of the sample to be tested.

[0048] The above mapping model has the following advantages: 1) Handling uncertainty: The diffusion model is inherently good at handling one-to-many mapping relationships. The same hyperspectral feature may correspond to a set of slightly different mass spectra (due to instrument errors, minute differences in samples, etc.). The diffusion model can generate a reasonable sample from all these possible results, rather than a fuzzy average. 2) High-quality generation: The diffusion model can produce very clear and realistic data. 3) Data completion capability: The model naturally has the ability to "complete" the data. If the mass spectrometry imaging data is partially missing (i.e., there is "known partial mass spectrometry information"), it can be used as a strong condition combined with the hyperspectral condition to guide the model to generate the missing part, thereby achieving data repair and enhancement.

[0049] In one possible implementation, after generating the molecular image of the sample to be tested in step 250, the method further includes step 260, determining whether the molecular image of the sample to be tested satisfies the iteration termination condition. The iteration termination condition may include at least one of the following: 1) The molecular image of the sample to be tested meets the imaging requirements.

[0050] For example, calculate the root mean square error (RMSE) between predicted and measured values ​​on a sparse calibration point set, where the RMSE is below a set threshold.

[0051] 2) The number of calibration points reaches the first preset threshold.

[0052] 3) The contribution of newly added calibration points to the optimization of the mapping model is lower than the second preset threshold.

[0053] For example, the improvement in RMSE after two consecutive iterations is less than the second preset threshold.

[0054] 4) The similarity between the optimized molecular image of the test sample and the benchmark molecular image reaches the third set threshold.

[0055] For example, multiple mass spectrometry imaging sites that were not involved in modeling are randomly selected to form a benchmark molecular image validation set, and the structural similarity index (SSIM) between the optimized molecular image of the test sample and the benchmark molecular image is determined to reach a third set threshold.

[0056] If the iteration termination condition is not met, proceed to steps 261 to 264 until the iteration termination condition is met; if the iteration termination condition is met, proceed to steps 265 and 266.

[0057] Step 261: Add the points in the molecular image of the sample to be tested whose prediction uncertainty is higher than the set threshold as new calibration points and add them to the current sparse calibration point set to obtain the expanded sparse calibration point set.

[0058] For example, the prediction variance of each pixel in the molecular image of the sample to be tested can be calculated, and the N pixels with the largest prediction variance can be selected as new calibration points to minimize model uncertainty with the fewest number of measurements. The new calibration points are then added to the current sparse calibration point set to obtain the expanded sparse calibration point set.

[0059] Step 262: Obtain mass spectrometry imaging data for the newly added calibration points.

[0060] To obtain mass spectrometry imaging data for the newly added calibration point, refer to step 230.

[0061] Step 263: Optimize the current mapping model based on the expanded sparse calibration point set to obtain the optimized mapping model.

[0062] The mapping model is retrained based on the expanded sparse calibration point set, and the optimized mapping model is obtained by referring to the above mapping model establishment process.

[0063] Step 264: Based on the optimized mapping model, calibrate the optical imaging data of the sample to be tested to generate an optimized molecular image of the sample to be tested.

[0064] Based on the optimized mapping model, the optical imaging data of the sample to be tested is calibrated again to generate an optimized molecular image of the sample to be tested.

[0065] Step 265: Output the optimized molecular image of the sample to be tested as the molecular image of the sample to be tested.

[0066] If the iteration termination condition is met, the optimized molecular image of the sample to be tested will be output as the molecular image of the sample to be tested. High-fidelity molecular images of the sample to be tested will be output, and users can use image analysis software (such as ImageJ) to obtain the quantitative spatial distribution information of characteristic molecules in any region of interest (such as tumor cell region, normal cell region), achieving accurate identification and molecular expression analysis of any region of the sample.

[0067] Step 266: Add the current mapping model and the relevant data of the optimized molecular image of the sample to be tested to the molecular imaging database as samples.

[0068] After completing the analysis of the sample, the system adds the final optimized mapping model (including model parameters and feature variables) and related data of the optimized molecular image of the sample (such as the mean of the hyperspectral feature spectrum of the sample and the relative content range of the main chemical components (such as phospholipids and proteins)) to the molecular imaging database.

[0069] In the above technical solution, in regions with high uncertainty in the mapping model prediction, additional sampling points for mass spectrometry imaging data are added. The current mapping model is iteratively optimized using the expanded sparse calibration point set, continuously improving the quantitative accuracy of optical imaging data and the spatial resolution of mass spectrometry imaging. This iteration is repeated until a preset termination condition is met, generating a high-fidelity molecular image that highly approximates the effect of mass spectrometry imaging in terms of molecular recognition accuracy and spatial detail. The spatial resolution of this image depends on the initial optical imaging technology, and the chemical specificity and quantitative accuracy reach a level comparable to mass spectrometry imaging after iterative calibration.

[0070] In one possible implementation, before step 220, which involves selecting multiple calibration points from the sample to be tested based on optical imaging data to obtain a sparse calibration point set, the following steps are also included: Step 211: Calculate the similarity between the optical imaging data of the sample to be tested and the optical imaging data of the samples in the molecular imaging database.

[0071] Step 212: Select the mapping model corresponding to the sample with the highest similarity as the initial mapping model.

[0072] Step 213: Based on the initial mapping model, calibrate the optical imaging data of the sample to be tested to generate the initial molecular image of the sample to be tested.

[0073] After acquiring the optical imaging data of the sample to be tested, the cosine similarity between the characteristic spectrum of the sample and the characteristic spectra of samples in the molecular imaging database can be calculated first. The mapping model corresponding to the sample with the highest similarity is selected as the initial mapping model. Then, based on the initial mapping model, the optical imaging data of the sample to be tested is calibrated to generate the initial molecular image of the sample.

[0074] Step 214: Determine whether the initial molecular image meets the imaging requirements.

[0075] If the initial molecular image meets the imaging requirements, proceed to step 215; if the initial molecular image does not meet the imaging requirements, proceed to step 216.

[0076] Step 215: Output the initial molecular image as the molecular image of the sample to be tested.

[0077] Step 216: Optimize the initial mapping model.

[0078] Optimize the initial mapping model, referring to steps 261 to 264.

[0079] If the initial molecular image does not meet the imaging requirements, the initial mapping model can be optimized based on the initial molecular image. Only a small number of calibration points need to be added for verification and fine-tuning. This can quickly complete the iterative calibration of the sample to be tested and generate a molecular image that meets the imaging requirements, which can significantly reduce the number of iterations and analysis time required for new samples to be tested.

[0080] In the above technical solution, an updatable molecular imaging database is constructed. The mapping model obtained from historical sample calibration is transferred as prior knowledge to the imaging analysis of the sample to be tested. Combined with an iterative mechanism of active learning, the system has self-optimization capabilities, and the calibration effect continuously improves with data accumulation, forming a closed-loop optimization. For similar series of samples (such as lung cancer tissue sections), the number of iterative calibrations for new samples to be tested can be greatly reduced, significantly improving the overall efficiency of the analysis sequence. This is particularly suitable for scenarios such as large-scale clinical pathological sample testing and large-scale industrial material composition screening.

[0081] Figure 4 A schematic diagram of the overall process of a multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration provided in this application embodiment is shown below. Figure 4 As shown, the method includes the following steps: Step 401: Obtain optical imaging data of the sample to be tested.

[0082] Step 402: Calculate the similarity between the optical imaging data of the sample to be tested and the optical imaging data of the samples in the molecular imaging database.

[0083] Step 403: Select the mapping model corresponding to the sample with the highest similarity as the initial mapping model.

[0084] Step 404: Based on the initial mapping model, calibrate the optical imaging data of the sample to be tested to generate an initial molecular image of the sample to be tested.

[0085] Step 405: Determine whether the initial molecular image meets the imaging requirements.

[0086] If the imaging requirements are not met, proceed to step 406; if the imaging requirements are met, proceed to step 412.

[0087] Step 406: Based on optical imaging data, select multiple calibration points from the sample to be tested to obtain a sparse calibration point set.

[0088] Step 407: Obtain mass spectrometry imaging data of each calibration point in the sparse calibration point set.

[0089] Step 408: Using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value, establish a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding points.

[0090] Step 409: Based on the mapping model, calibrate the optical imaging data of the sample to be tested to generate a molecular image of the sample to be tested.

[0091] Step 410: Determine whether the iteration termination condition is met.

[0092] If the iteration termination condition is not met, proceed to step 411; if the iteration termination condition is met, proceed to step 412.

[0093] Step 411: Add the points in the molecular image of the sample to be tested that have a prediction uncertainty higher than the set threshold as new calibration points to the current sparse calibration point set.

[0094] Step 412: Output the current molecular image.

[0095] Step 413: Add the current mapping model and related data of molecular images as samples to the molecular imaging database.

[0096] This application provides a multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration. Using sparse mass spectrometry data as the baseline, a mapping model is constructed and iteratively optimized through machine learning algorithms to achieve precise calibration of optical imaging data and build a molecular imaging database for knowledge transfer. The resulting high-fidelity molecular images maintain the high spatial resolution of the initial optical imaging while achieving chemical specificity and quantitative accuracy comparable to mass spectrometry imaging through dynamic iterative calibration. This method addresses the problems of low quantitative accuracy in traditional optical imaging and low throughput in mass spectrometry imaging, as well as spatial resolution limited by sampling density. It can be widely applied in clinical pathology analysis, material composition detection, and environmental pollutant distribution imaging, and is compatible with various probe ionization techniques that meet specific standards.

[0097] The following describes the multimodal molecular imaging device based on DESI mass spectrometry dynamic iterative calibration provided by the present invention. The multimodal molecular imaging device based on DESI mass spectrometry dynamic iterative calibration described below can be referred to in correspondence with the multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration described above.

[0098] Figure 5 A schematic diagram of a multimodal molecular imaging device based on DESI mass spectrometry dynamic iterative calibration provided in this application embodiment is shown below. Figure 5As shown, the device 500 includes: The acquisition module 510 is used to acquire optical imaging data of the sample to be tested; Processing module 520 is used to select multiple calibration points from the sample to be tested based on the optical imaging data to obtain a sparse calibration point set; The acquisition module 510 is also used to acquire mass spectrometry imaging data of each calibration point in the sparse calibration point set; The mapping model module 530 is used to establish a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding points, using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value. The generation module 540 is used to calibrate the optical imaging data of the sample under test based on the mapping model, and generate a molecular image of the sample under test.

[0099] Optionally, the processing module 520 is further configured to determine whether the molecular image of the sample to be tested satisfies the iteration termination condition; If the iteration termination condition is not met, the following steps are performed until the iteration termination condition is met: Points in the molecular image of the sample to be tested with prediction uncertainty higher than a set threshold are added as new calibration points and added to the current sparse calibration point set to obtain an expanded sparse calibration point set. The acquisition module 510 is also used to acquire the mass spectrometry imaging data of the newly added calibration point; The mapping model module 530 is also used to optimize the current mapping model based on the expanded sparse calibration point set to obtain an optimized mapping model. The generation module 540 is also used to calibrate the optical imaging data of the sample to be tested based on the optimized mapping model, and generate an optimized molecular image of the sample to be tested.

[0100] Optionally, the processing module 520 is further configured to output the optimized molecular image of the sample to be tested as the molecular image of the sample to be tested if the iteration termination condition is met. The current mapping model and the relevant data of the optimized molecular images of the test samples are added to the molecular imaging database as samples.

[0101] Optionally, before selecting multiple calibration points from the sample to be tested based on the optical imaging data to obtain a sparse calibration point set, the processing module 520 is further used to calculate the similarity between the optical imaging data of the sample to be tested and the optical imaging data of the samples in the molecular imaging database. The mapping model corresponding to the sample with the highest similarity is selected as the initial mapping model; The generation module 540 is also used to calibrate the optical imaging data of the sample to be tested based on the initial mapping model, and generate an initial molecular image of the sample to be tested. If the initial molecular image meets the imaging requirements, the processing module 520 is further configured to output the initial molecular image as the molecular image of the sample to be tested. If the initial molecular image does not meet the imaging requirements, the mapping model module 530 is further used to optimize the initial mapping model.

[0102] Optionally, the mapping model module 530 is further configured to establish a mapping model using the following method: The optical imaging data of each calibration point of the sample under test is used to generate noise data for each calibration point by adding noise to the optical imaging data of each calibration point through the forward generation module. The noise data, location information, and cell clustering code of each calibration point are input into the inverse noise reduction module to train the neural network in the inverse noise reduction module, so that the neural network can generate mass spectrometry imaging data corresponding to each calibration point of the sample to be tested. The neural network is used as the mapping model.

[0103] Optionally, the iteration termination condition includes at least one of the following: The molecular images of the sample to be tested meet the imaging requirements; The number of calibration points has reached the first preset threshold. The contribution of newly added calibration points to the optimization of the mapping model is less than the second preset threshold; The similarity between the optimized molecular image of the test sample and the baseline molecular image reaches the third set threshold.

[0104] It should be noted that the molecular imaging device provided in this application embodiment can realize all the method steps implemented in the above-mentioned multimodal molecular imaging method embodiment based on DESI mass spectrometry dynamic iterative calibration, and can achieve the same technical effect. Therefore, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail here.

[0105] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communication bus 640. The processor 610, communications interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions from the memory 630 to execute the multimodal molecular imaging methods based on DESI mass spectrometry dynamic iterative calibration provided by the methods described above.

[0106] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0107] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to perform the molecular imaging provided by the methods described above.

[0108] In another aspect, this application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the molecular imaging provided by the methods described above.

[0109] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration, characterized in that, The method includes: Acquire optical imaging data of the sample to be tested; Based on the optical imaging data, a sparse calibration point set is obtained by selecting multiple calibration points from the sample to be tested. Obtain mass spectrometry imaging data of each calibration point in the sparse calibration point set; Using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value, a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding point is established. Based on the mapping model, the optical imaging data of the sample to be tested is calibrated to generate a molecular image of the sample to be tested.

2. The multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration according to claim 1, characterized in that, The method further includes: Determine whether the molecular image of the sample to be tested meets the iteration termination condition; If the iteration termination condition is not met, the following steps are performed until the iteration termination condition is met: Points in the molecular image of the sample to be tested with prediction uncertainty higher than a set threshold are added as new calibration points and added to the current sparse calibration point set to obtain an expanded sparse calibration point set. Acquire the mass spectrometry imaging data of the newly added calibration points; The current mapping model is optimized based on the expanded sparse calibration point set to obtain the optimized mapping model. Based on the optimized mapping model, the optical imaging data of the sample to be tested is calibrated to generate an optimized molecular image of the sample to be tested.

3. The multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration according to claim 2, characterized in that, The method further includes: If the iteration termination condition is met, the optimized molecular image of the sample to be tested will be output as the molecular image of the sample to be tested. The current mapping model and the relevant data of the optimized molecular images of the test samples are added to the molecular imaging database as samples.

4. The multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration according to claim 3, characterized in that, Before selecting multiple calibration points from the sample under test based on the optical imaging data to obtain a sparse calibration point set, the method further includes: Calculate the similarity between the optical imaging data of the sample to be tested and the optical imaging data of the samples in the molecular imaging database; The mapping model corresponding to the sample with the highest similarity is selected as the initial mapping model; Based on the initial mapping model, the optical imaging data of the sample to be tested is calibrated to generate an initial molecular image of the sample to be tested. If the initial molecular image meets the imaging requirements, then the initial molecular image is output as the molecular image of the sample to be tested; If the initial molecular image does not meet the imaging requirements, the initial mapping model is optimized.

5. The multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration according to any one of claims 1 to 4, characterized in that, The method further includes: The mapping model is established using the following method: The optical imaging data of each calibration point of the sample under test is used to generate noise data for each calibration point by adding noise to the optical imaging data of each calibration point through the forward generation module. The noise data, location information, and cell clustering code of each calibration point are input into the inverse noise reduction module to train the neural network in the inverse noise reduction module, so that the neural network can generate mass spectrometry imaging data corresponding to each calibration point of the sample to be tested. The neural network is used as the mapping model.

6. The multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration according to claim 2, characterized in that, The iteration termination condition includes at least one of the following: The molecular images of the sample to be tested meet the imaging requirements; The number of calibration points has reached the first preset threshold. The contribution of newly added calibration points to the optimization of the mapping model is less than the second preset threshold; The similarity between the optimized molecular image of the test sample and the baseline molecular image reaches the third set threshold.

7. A molecular imaging device, characterized in that, The device includes: The acquisition module is used to acquire optical imaging data of the sample under test. The processing module is used to select multiple calibration points from the sample under test based on the optical imaging data to obtain a sparse calibration point set; The acquisition module is also used to acquire mass spectrometry imaging data of each calibration point in the sparse calibration point set; The mapping model module is used to establish a mapping model between the mass spectrometry imaging data and the optical imaging data of the corresponding points, using the mass spectrometry imaging data of each calibration point in the sparse calibration point set as the reference true value. The generation module is used to calibrate the optical imaging data of the sample under test based on the mapping model, and generate a molecular image of the sample under test.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the multimodal molecular imaging method based on DESI mass spectrometry dynamic iterative calibration as described in any one of claims 1 to 6.

Images