Multi-modal image fusion breast cancer screening analysis system and method

Abstract

The invention relates to the technical field of image analysis, in particular to a breast cancer screening analysis system and method based on multi-modal image fusion, and the system comprises an information entropy weight adjustment module, a time-frequency domain feature processing module, a geometric correction fusion module, a data real-time updating module and a comprehensive analysis output module. According to the method, weight distribution is optimized by evaluating the contribution degree of each mode in a specific scene, the data processing accuracy in the diagnosis process is enhanced, time-frequency domain conversion and filtering synchronization are carried out on image data, efficient integration of different mode features is achieved, and image alignment is more accurate through multi-scale decomposition and feature matching of geometric correction; the method has the advantages of reducing the misdiagnosis rate, flexibly adjusting and keeping the timeliness and adaptability of a processing strategy according to new diagnosis data, and remarkably improving the accuracy of early detection of breast cancer by strengthening key diagnosis features and improving the expression quality of image data, thereby playing a positive role in improving the survival rate and life quality of patients.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a multi-modal image fusion breast cancer screening analysis system and method. Background Art

[0002] Image analysis is a broad technical field that focuses on processing and parsing image content to extract useful information. This field combines image processing, computer vision, pattern recognition, and machine learning technologies to enhance, restore, compress, and describe images. In the medical field, image analysis technology is particularly important. It supports disease diagnosis, treatment planning, and efficacy monitoring by analyzing medical imaging data (such as X-rays, CT, MRI, and ultrasound images). It can automatically identify and quantify structures, tissue types, and pathological changes in images, greatly improving the accuracy and efficiency of diagnosis.

[0003] Among them, the breast cancer screening and analysis system is an important application of image analysis technology in the medical field. The system uses image data, combined with advanced image processing and machine learning algorithms, to detect and diagnose breast cancer at an early stage. The system analyzes breast X-ray imaging (such as mammal photography), MRI or ultrasound images to identify possible tumors or abnormal structures, thereby helping doctors make accurate diagnoses. Its main purpose is to increase the early detection rate of breast cancer, optimize the screening process, and reduce unnecessary biopsies, thereby improving patients' survival rate and quality of life, which is of great significance to improving public health.

[0004] Existing image analysis technologies often show insufficient flexibility and poor adaptability in multimodal data processing. Especially in terms of data integration, various imaging technologies cannot be effectively coordinated and their respective advantages cannot be fully utilized, which affects the comprehensiveness and accuracy of diagnosis. For example, in single-modality analysis, the risk of misdiagnosis or missed diagnosis may increase due to inaccurate recognition. In addition, many existing systems fail to update diagnostic data in real time and lack the ability to quickly adapt to the latest clinical information, which is particularly disadvantageous in medical environments and limits the potential of image analysis in improving diagnostic efficiency and reducing unnecessary surgery. These technical limitations not only affect the formulation of treatment plans, but may also have a negative impact on patients' quality of life. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a breast cancer screening and analysis system and method based on multimodal image fusion.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: A multimodal image fusion breast cancer screening and analysis system comprises:

[0007] The information entropy weight adjustment module collects X-ray, MRI, and ultrasound image data, calculates the information entropy of all modal image data, evaluates the information contribution of each modality in the current breast cancer screening scenario, dynamically adjusts the weight according to the information gain comparison results, and generates information entropy weight results;

[0008] The time-frequency domain feature processing module performs fast Fourier transform on the information entropy weight result, analyzes the features in the time-frequency domain, synchronizes the difference modal features through frequency domain filtering, fuses the frequency domain features, and restores the data to the time domain through inverse Fourier transform to generate frequency domain feature synchronization results;

[0009] The geometric correction fusion module performs multi-scale decomposition and feature matching on the images of the difference modes based on the frequency domain feature synchronization result, uses spatial correction to adjust the image alignment and performs fusion processing, performs quality assessment, and generates a fusion quality assessment result;

[0010] The data real-time update module collects diagnostic data in real time, updates the modality weight and correction strategy according to the newly collected diagnostic data and the fusion quality assessment results, adjusts the image processing flow to match the actual conditions of the patients with differences, and generates real-time update results;

[0011] The comprehensive analysis output module summarizes the real-time update results, performs detail expression optimization, analyzes the image data of each modality, identifies and enhances key diagnostic features, and outputs comprehensive screening analysis results.

[0012] As a further scheme of the present invention, the information entropy weight results include weight parameters, entropy values ​​and information gain analysis results, the frequency domain feature synchronization results include frequency domain graphs, time recovery data and filtering features, the fusion quality assessment results include correction parameters, fusion metrics and alignment accuracy, the real-time update results include weight adjustment records, processing parameters and data update logs, and the comprehensive screening and analysis results include image quality scores, feature recognition records and data comprehensive records.

[0013] As a further solution of the present invention, the information entropy weight adjustment module includes a data acquisition submodule, an information entropy calculation submodule, and a weight adjustment submodule.

[0014] The data acquisition submodule retrieves and collects X-ray, MRI, and ultrasound multimodal image data from the medical image database according to the needs of breast cancer screening, performs standardized processing on statistical data quality and format, and obtains a multimodal image dataset;

[0015] The information entropy calculation submodule performs weighted calculation on the information entropy of X-ray, MRI and ultrasound images through the multimodal image data set, quantitatively evaluates the information contribution of the modalities, and obtains the information entropy evaluation result;

[0016] The weight adjustment submodule dynamically adjusts the weight of the modal image data based on the information entropy evaluation result, performs information gain evaluation and reallocates weight parameters to form an information entropy weight result.

[0017] As a further solution of the present invention, the information entropy is weightedly calculated according to the formula:

[0018]

[0019] Calculate the information contribution E of the modal image, where H(X i ) represents the information entropy of the i-th modality image, which is used to measure the uncertainty of the information contained in the image. M represents the number of different modalities, which is used to define the total number of modalities considered in the evaluation process.

[0020] As a further solution of the present invention, the time-frequency domain feature processing module includes a fast Fourier transform submodule, a frequency domain feature fusion submodule, and an inverse Fourier transform submodule.

[0021] The fast Fourier transform submodule performs fast Fourier transform based on the information entropy weight result, converts the time series data into frequency domain data, analyzes the frequency points to extract amplitude and phase information, calculates the energy distribution of each frequency point, and obtains the frequency domain signal characteristics;

[0022] The frequency domain feature fusion submodule adopts the frequency domain signal features to perform multimodal frequency domain filtering, comprehensively processes the frequency domain information of X-ray, MRI and ultrasound image data, synchronously adjusts the feature differences between the modalities, confirms the signal synchronization, and obtains synchronous frequency domain features;

[0023] The inverse Fourier transform submodule applies inverse Fourier transform through the synchronized frequency domain features, restores the frequency domain feature data to the time domain format, verifies the integrity and practicality of the data, reconstructs the time domain image data, and generates the frequency domain feature synchronization results.

[0024] As a further solution of the present invention, the energy distribution of each frequency point is according to the formula:

[0025]

[0026] Calculation is performed, where P(f) represents the energy distribution value of each frequency point, N represents the total number of samples of the time series data, represents the number of points in the signal, X(n) represents the signal amplitude of the nth point in the time series, the original data is provided directly, f represents the target frequency, from 0 to N-1, and each frequency point corresponds to the calculation of an energy value. It is an adjustment factor used to reduce the weight of high-frequency signals in energy distribution.

[0027] As a further solution of the present invention, the geometric correction fusion module includes a multi-scale decomposition submodule, a feature matching submodule, and a fusion quality assessment submodule.

[0028] The multi-scale decomposition submodule performs multi-scale decomposition of the image based on the frequency domain feature synchronization result, separates the feature layers of difference scale and direction for the image data of the difference mode, analyzes the feature intensity of each scale layer, extracts key edge and texture information, and obtains the decomposed feature layer;

[0029] The feature matching submodule adopts the decomposed feature layer, determines the feature matching points through pattern recognition, performs spatial alignment, corrects the displacement and rotation differences between images, optimizes the matching accuracy between images, and obtains the corrected alignment result;

[0030] The fusion quality assessment submodule integrates the difference modality images through the correction alignment results, performs quality analysis on the fused images, evaluates the clarity, contrast and detail retention, and generates a fusion quality assessment result.

[0031] As a further solution of the present invention, the real-time data update module includes a data collection submodule, a weight update submodule, and a process adjustment submodule.

[0032] The data collection submodule directly receives the latest imaging data and physiological measurement data from the diagnostic equipment through real-time monitoring, performs continuous stream input of data, performs comprehensive information update, and obtains the latest diagnostic data set;

[0033] The weight update submodule analyzes the impact of new data on existing modality weights through the latest diagnostic data set and the fusion quality assessment results, uses analytical statistics to update and adjust weights, reveals the actual requirements of the latest diagnostic information for image processing, and outputs updated weight parameters;

[0034] The process adjustment submodule reconfigures the image processing process based on the updated weight parameters, performs process adjustment optimization on each modality image, and generates real-time updated results.

[0035] As a further solution of the present invention, the comprehensive analysis output module includes a result summary submodule, a feature enhancement submodule, and a comprehensive output submodule.

[0036] The result summary submodule integrates the multimodal diagnostic data through data synchronization based on the real-time update results, unifies the data format and timestamp, checks the consistency and integrity of the data, and generates a standardized diagnostic data set;

[0037] The feature enhancement submodule enhances the diagnostic features through the standardized diagnostic data set, applies local contrast enhancement and edge sharpening to key areas of the image, enhances the diagnostic information in the image, and outputs the feature enhancement result;

[0038] The comprehensive output submodule comprehensively analyzes all images and data based on the feature enhancement results, performs physician review and quality control on the analysis results, determines the accuracy and reliability of the analysis, and outputs the comprehensive screening analysis results.

[0039] A multimodal image fusion breast cancer screening and analysis method comprises the following steps:

[0040] S1: According to the needs of breast cancer screening, retrieve and collect X-ray, MRI and ultrasound multimodal image data, calculate the image information entropy, and quantitatively evaluate the information contribution of the modality to obtain the information entropy evaluation result;

[0041] S2: Based on the information entropy evaluation result, dynamically adjust the weight of the modal image data, perform fast Fourier transform, analyze the frequency points to extract amplitude and phase information, calculate the energy distribution of each frequency point, and obtain frequency domain signal characteristics;

[0042] S3: using the frequency domain signal features, performing multimodal frequency domain filtering, synchronously adjusting the feature differences between the modes, applying inverse Fourier transform, reconstructing the time domain image data, and generating frequency domain feature synchronization results;

[0043] S4: Based on the frequency domain feature synchronization result, perform multi-scale decomposition of the image, analyze the feature strength of each scale layer, determine the feature matching points to perform spatial alignment, perform quality analysis on the fused image, and generate a fusion quality assessment result;

[0044] S5: receiving the latest imaging data and physiological measurement data, combining the fusion quality assessment results, analyzing the impact of the new data on the existing modality weights, updating and adjusting the weights, reconfiguring the image processing flow, and generating real-time update results;

[0045] S6: Based on the real-time update results, integrate multimodal diagnostic data, unify data formats and timestamps, enhance diagnostic features, conduct physician review and quality control on analysis results, determine analysis accuracy and reliability, and output comprehensive screening analysis results.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are:

[0047] In the present invention, the contribution of each mode in a specific scenario is evaluated to optimize the weight distribution, enhance the accuracy of data processing in the diagnosis process, perform time-frequency domain conversion and filter synchronization on the image data, achieve efficient integration of different modal features, multi-scale decomposition and feature matching of geometric correction to make image alignment more accurate and reduce the misdiagnosis rate. Flexible adjustments are made according to new diagnostic data to maintain the timeliness and adaptability of the processing strategy. By strengthening key diagnostic features and improving the expression quality of image data, the accuracy of early detection of breast cancer is significantly improved, thereby playing a positive role in improving the survival rate and quality of life of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a system flow chart of the present invention;

[0049] Figure 2 This is a flow chart of the information entropy weight adjustment module of the present invention;

[0050] Figure 3 This is a flow chart of the time-frequency domain feature processing module of the present invention;

[0051] Figure 4 This is a flow chart of the geometric correction fusion module of the present invention;

[0052] Figure 5 This is a flow chart of the real-time data update module of the present invention;

[0053] Figure 6 It is a flow chart of the comprehensive analysis output module of the present invention;

[0054] Figure 7 The figure is a flow chart of the steps of the method of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0057] See also Figure 1 , a multimodal image fusion breast cancer screening and analysis system includes:

[0058] The information entropy weight adjustment module collects X-ray, MRI, and ultrasound image data, calculates the information entropy of all modal image data, evaluates the information contribution of each modality in the current breast cancer screening scenario, dynamically adjusts the weight according to the information gain comparison results, and generates information entropy weight results;

[0059] The time-frequency domain feature processing module performs fast Fourier transform on the information entropy weight results, analyzes the features in the time-frequency domain, synchronizes the difference modal features through frequency domain filtering, fuses the frequency domain features, and restores the data to the time domain through inverse Fourier transform to generate frequency domain feature synchronization results;

[0060] The geometric correction fusion module performs multi-scale decomposition and feature matching on the difference modal images based on the frequency domain feature synchronization results, uses spatial correction to adjust the image alignment and performs fusion processing, performs quality assessment, and generates fusion quality assessment results;

[0061] The data real-time update module collects diagnostic data in real time, updates the modality weight and correction strategy based on the newly collected diagnostic data and fusion quality assessment results, adjusts the image processing process to match the actual conditions of different patients, and generates real-time update results;

[0062] The comprehensive analysis output module summarizes the real-time update results, performs detail expression optimization, analyzes the image data of each modality, identifies and enhances key diagnostic features, and outputs comprehensive screening analysis results.

[0063] The information entropy weight results include weight parameters, entropy values ​​and information gain analysis results; the frequency domain feature synchronization results include frequency domain graphs, time recovery data and filtering features; the fusion quality assessment results include correction parameters, fusion metrics and alignment accuracy; the real-time update results include weight adjustment records, processing parameters and data update logs; the comprehensive screening and analysis results include image quality scores, feature recognition records and data synthesis records.

[0064] See also Figure 2 The information entropy weight adjustment module includes a data acquisition submodule, an information entropy calculation submodule, and a weight adjustment submodule.

[0065] The data acquisition submodule retrieves and collects X-ray, MRI, and ultrasound multimodal image data from the medical image database according to the needs of breast cancer screening, performs standardized processing on statistical data quality and format, and obtains a multimodal image dataset;

[0066] The data acquisition submodule first selects X-ray, MRI and ultrasound image data that are suitable for breast cancer screening from the medical image database according to the query parameters. The system sends specific parameters such as image type, case number and image acquisition date through API calls, performs preliminary screening on the returned image data, and excludes data with unqualified image quality or unclear markings. At the same time, it adjusts the image resolution and contrast according to the set standardized process to ensure the consistency of subsequent processing, converts and compresses the image format to reduce storage space and increase data processing speed, counts the number and type of processed data, verifies the integrity and accuracy of the data set, and forms the final multimodal image data set for subsequent information entropy calculation and weight adjustment analysis.

[0067] The information entropy calculation submodule uses a multimodal image data set to perform weighted calculations on the information entropy of X-ray, MRI, and ultrasound images, and quantitatively evaluates the information contribution of the modalities to obtain information entropy evaluation results.

[0068] Information entropy is weighted and calculated according to the formula:

[0069]

[0070] Calculate the information contribution of the modal image, where H(X i ) represents the information entropy of the i-th modality image, which is used to measure the uncertainty of the information contained in the image. M represents the number of different modalities, which is used to define the total number of modalities considered in the evaluation process.

[0071] There are three modalities: X-ray, MRI and ultrasound images, so M = 3. i ) is the information entropy of the i-th modality image. Information entropy H(X i ) is calculated by monitoring the grayscale distribution or other image characteristics of each modality image, and the information entropy is obtained by summing up the information of each pixel and taking the average.

[0072] The information entropy values ​​of each modality are collected as follows:

[0073] H(X1)(X-ray image information entropy)=2.3

[0074] H(X2)(MRI image information entropy)=3.1

[0075] H(X3)(ultrasound image information entropy)=1.8

[0076] Calculate the sum of the absolute values ​​of each modal information entropy:

[0077]

[0078] Then calculate the average of the sum:

[0079]

[0080] Finally, calculate the square root of the average value to get the final evaluation value E:

[0081]

[0082] The results show that the average information entropy of the entire multimodal image dataset is approximately 1.55 after adjustment, and the value reflects the average amount of information in the image dataset after integrating different modalities.

[0083] The weight adjustment submodule dynamically adjusts the weight of the modal image data based on the information entropy evaluation results, performs information gain evaluation and reallocates weight parameters to form information entropy weight results;

[0084] According to the information entropy evaluation results, the weight adjustment submodule first analyzes the information entropy value of each modality image data, and dynamically adjusts the initial weight according to the information entropy size of each modality. During the adjustment process, based on the principle of maximizing the difference, the optimal weight adjustment factor is determined. The adjustment factor reflects the importance of different modality data in the overall data set. The adjusted weight is applied to each modality through linear combination to enhance the influence of the modality with larger information content. After multiple rounds of iterative optimization, the information gain is evaluated to ensure that each adjustment effectively improves the overall information value of the data. The final information entropy weight result provides accurate data input for the breast cancer screening model, in order to improve the diagnostic efficiency and accuracy of the model.

[0085] See also Figure 3 The time-frequency domain feature processing module includes a fast Fourier transform submodule, a frequency domain feature fusion submodule, and an inverse Fourier transform submodule.

[0086] The fast Fourier transform submodule performs fast Fourier transform based on the information entropy weight result, converts the time series data into frequency domain data, analyzes the frequency points to extract the amplitude and phase information, calculates the energy distribution of each frequency point, and obtains the frequency domain signal characteristics;

[0087] The energy distribution of each frequency point is based on the formula:

[0088]

[0089] Calculation is performed, where P(f) represents the energy distribution value of each frequency point, N represents the total number of samples of the time series data, represents the number of points in the signal, X(n) represents the signal amplitude of the nth point in the time series, the original data is provided directly, f represents the target frequency, from 0 to N-1, and each frequency point corresponds to the calculation of an energy value. It is an adjustment factor used to reduce the weight of high-frequency signals in energy distribution.

[0090] In one signal acquisition, a total of N = 1000 data points are collected. X(n) is the signal amplitude measured at the nth time point. Consider f is a specific frequency, from f = 0 to f = 999.

[0091] Assume that the measured value of X(n) when n=0 is X(0)=2, and the measured value when n=1 is X(1)=3. Assume f=0 and start the calculation.

[0092] Calculate P(0) when f = 0:

[0093]

[0094] The results show that the signal energy at frequency f = 0 is very low, corresponding to the energy level of the DC component. Through such calculations, the energy distribution of the entire spectrum can be analyzed to understand the relative importance of different frequency components in the signal. By adjusting the factors to ensure a more reasonable evaluation of low-frequency signals, the frequency domain analysis of the entire signal is more consistent with the physical characteristics of the actual signal. It can not only effectively reduce the interference of high-frequency noise, but also balance the energy distribution of the entire frequency domain, providing more accurate and practical data support for signal processing.

[0095] The frequency domain feature fusion submodule uses frequency domain signal features to perform multimodal frequency domain filtering, comprehensively processes the frequency domain information of X-ray, MRI and ultrasound image data, synchronously adjusts the feature differences between modalities, confirms signal synchronization, and obtains synchronous frequency domain features;

[0096] The frequency domain feature fusion submodule uses frequency domain signal features. First, multimodal frequency domain filtering is performed on the frequency domain signals of X-ray, MRI and ultrasound image data, including the use of bandpass filters to remove noise and enhance signals in specific frequency bands. At the same time, the frequency domain features of different modes are linearly or nonlinearly combined to ensure that the extracted features can reflect the comprehensive information of each mode. By adjusting the filtering parameters and combination weights, the feature differences between the modes are synchronously adjusted, and the contribution of each mode is further analyzed and their synchronization is optimized. In this process, the integration of frequency domain features ensures the consistency and synchronization of signals between different modes, and finally obtains synchronized frequency domain features that can accurately describe the comprehensive characteristics of multimodal data, preparing accurate input for data recovery of the inverse Fourier transform submodule.

[0097] The inverse Fourier transform submodule restores the frequency domain feature data to the time domain format by synchronizing the frequency domain features and applying the inverse Fourier transform, verifies the integrity and practicality of the data, reconstructs the time domain image data, and generates the frequency domain feature synchronization results;

[0098] The inverse Fourier transform submodule processes the synchronized frequency domain features obtained by the frequency domain feature fusion submodule. First, the inverse Fourier transform (IFFT) algorithm is applied to convert these frequency domain data back to the time domain format. This operation involves recombining the amplitude and phase information of each frequency point into the complex form of the time series, and then calculating the inverse transform of these complex numbers to reconstruct the time series of the original signal. This process ensures the accurate recovery of each data point, and verifies the integrity and practicality of the data by comparing the correlation between the original time series and the reconstructed sequence. Through multiple iterative adjustments, the inverse transform parameter optimization ensures that the reconstructed data maintains high quality. The final generated time domain image data will be used for further analysis and diagnostic processing. After this stage is completed, the specific results of frequency domain feature synchronization are formed, which provides an effective data basis for subsequent image processing and feature analysis.

[0099] See also Figure 4 ,The geometric correction fusion module includes a multi-scale decomposition submodule, a feature matching submodule, and a fusion quality assessment submodule.

[0100] The multi-scale decomposition submodule performs multi-scale decomposition of the image based on the frequency domain feature synchronization result, separates the feature layers of difference scale and direction for the image data of the difference mode, analyzes the feature intensity of each scale layer, extracts key edge and texture information, and obtains the decomposed feature layer;

[0101] The multiscale decomposition submodule performs multiscale decomposition based on the frequency domain feature synchronization results. By applying wavelet transforms of different scales to each modal image data, the image features of each scale level are distinguished and separated, and the feature intensity is determined for each scale layer. After that, special attention is paid to the key edge and texture information in the high-energy scale layer to extract the most significant image features. The process is completed through quantitative analysis to ensure that the features of each decomposition level are accurately identified and recorded. The processed decomposition feature layer includes the key visual and structural information of different modal images, which provides the necessary data basis for subsequent feature matching and image fusion. The final decomposition feature layer will be used for further pattern recognition and image alignment.

[0102] The feature matching submodule uses the decomposed feature layer to determine the feature matching points through pattern recognition, performs spatial alignment, corrects the displacement and rotation differences between images, optimizes the matching accuracy between images, and obtains the corrected alignment results;

[0103] The feature matching submodule uses the decomposed feature layer obtained by multi-scale decomposition, and begins to identify and calibrate the feature matching points in each modality image through pattern recognition technology. First, feature detection algorithms such as SIFT or SURF are used to locate the key points in the image, and then a unique representation is generated for each key point through feature descriptors to support matching between different images. Subsequently, a spatial alignment operation is performed, using transformation models such as affine transformation or perspective transformation to correct the displacement and rotation between images, ensuring that the feature points of different modality images accurately correspond in space, thereby optimizing the matching accuracy between images. After a series of processing, the system outputs the correction and alignment results, showing the synchronization and consistency of the image data of each modality at the feature level, providing an accurate alignment basis for subsequent image fusion and analysis.

[0104] The fusion quality assessment submodule corrects the alignment results, integrates the difference modality images, performs quality analysis on the fused images, evaluates the clarity, contrast, and detail retention, and generates fusion quality assessment results;

[0105] The fusion quality assessment submodule receives the corrected and aligned image results and starts to perform quality analysis on the fused data of the difference modality images. The process includes evaluating the clarity, contrast and detail retention of the fused image. First, image quality assessment is performed, such as the structural similarity index (SSIM) or contrast sensitivity function, to quantitatively analyze the consistency of the visual quality between the fused image and the original input image. Then, the difference between the fused image and the single modality image at the detail level is compared to evaluate the retention of detail information during the fusion process. In addition, the system also performs manual visual evaluation on the fused image, and experts score the fusion effect to ensure the practical application value of image fusion. Finally, the fusion quality assessment results are generated, which record the scores of various evaluation indicators in detail and provide a basis for further improving image fusion technology.

[0106] See also Figure 5 The real-time data update module includes a data collection submodule, a weight update submodule, and a process adjustment submodule.

[0107] The data collection submodule directly receives the latest imaging data and physiological measurement data from the diagnostic equipment through real-time monitoring, performs continuous stream input of data, performs comprehensive information update, and obtains the latest diagnostic data set;

[0108] The data collection submodule is connected to various diagnostic devices through a real-time monitoring system to directly receive the latest imaging data and physiological measurement data from the diagnostic devices. The process configures data flow management to support continuous input of high-speed data. The buffering mechanism within the system ensures that data will not be lost during transmission. At the same time, data standardization and preliminary filtering are implemented to exclude data input that is obviously erroneous or does not meet the standards. During the update process, the real-time monitoring system checks the data integrity and latest status according to the preset update frequency to ensure that the received data is the latest. In addition, it is connected to the data processing system through an interface to transmit the sorted data set for subsequent image processing and analysis. The latest diagnostic data set is obtained after a comprehensive update and will be directly used in the next stage of image processing, ensuring the timeliness and relevance of the processed data.

[0109] The weight update submodule uses the latest diagnostic data set and fusion quality assessment results to analyze the impact of new data on existing modality weights, uses analytical statistics to update and adjust weights, reveals the actual requirements of the latest diagnostic information for image processing, and outputs updated weight parameters;

[0110] After receiving the latest diagnostic data set and fusion quality assessment results, the weight update submodule analyzes and adjusts the weights. By comparing and analyzing the changes in the new and old data sets, the module identifies which modalities have increased or decreased in influence. Then, it applies dynamic weight adjustment to adjust the weights of each modality data based on the newly obtained statistical data and fusion quality assessment results. The adjustment is to optimize the overall performance of image processing. After the weight adjustment, the module outputs the updated weight parameters. The parameters reflect the impact of the latest diagnostic information on the image processing process, ensuring that the image processing system can dynamically optimize the processing strategy based on the latest data in order to more accurately respond to various diagnostic scenarios.

[0111] The process adjustment submodule reconfigures the image processing process based on the updated weight parameters, adjusts and optimizes the process for each modality image, and generates real-time updated results;

[0112] The process adjustment submodule starts to reconfigure and optimize the image processing process based on the updated weight parameters. The process includes resetting the parameters of each modality image processing, such as contrast enhancement, edge detection sensitivity, and noise filtering level, automatically analyzing the characteristics of each modality image data and the latest weight parameters, and determining the most suitable processing process. After adjustment, the system performs a simulation run test to ensure that the modified process can effectively process image data of various modalities, while ensuring that the image quality and processing speed meet clinical needs. After multiple iterative optimizations, real-time updated results are generated to display the new processes and expected outputs of various modal image processing, ensuring that the image processing system can quickly adapt and provide high-quality output when receiving new diagnostic data.

[0113] See also Figure 6 The comprehensive analysis output module includes the result summary submodule, the feature enhancement submodule, and the comprehensive output submodule.

[0114] The result summary submodule is based on real-time updated results, integrates multimodal diagnostic data through data synchronization, unifies data formats and timestamps, checks data consistency and integrity, and generates standardized diagnostic data sets;

[0115] After the result summary submodule is started, the multimodal diagnostic data are integrated through efficient data synchronization technology to unify the format and timestamp of each data source. In this process, the consistency and integrity of the data are automatically detected, and the missing or abnormal timestamps in the data are identified and corrected to ensure the uniformity of all data in time and format. The integrated data are standardized to meet the needs of subsequent processing procedures and a standardized diagnostic data set is generated. This data set ensures that data from different diagnostic devices and modes can be effectively used in subsequent analysis and processing, and provides an accurate data basis for ensuring data quality and consistency, which is used for further image enhancement and feature analysis processing.

[0116] The feature enhancement submodule enhances the diagnostic features by standardizing the diagnostic data set, applies local contrast enhancement and edge sharpening to the key areas of the image, strengthens the diagnostic information in the image, and outputs the feature enhancement results;

[0117] After receiving the standardized diagnostic data set, the feature enhancement submodule begins to enhance the diagnostic features in the image. The process includes applying local contrast enhancement technology and edge sharpening algorithm. Local contrast enhancement is achieved by adjusting the brightness and contrast parameters of specific areas in the image to make the details of the key areas more prominent. Edge sharpening is performed by enhancing the pixel gradient differences at the edges of objects in the image. All operations are based on precise calculations and preset thresholds to ensure that only the key parts of the diagnostic information are enhanced. The processed image features become clearer and more detailed due to the enhanced processing, which improves the accuracy and reliability of the diagnosis. The feature enhancement results are output, showing the processed key areas and details in the image, providing higher quality image data for the final diagnosis.

[0118] The comprehensive output submodule comprehensively analyzes all images and data based on the feature enhancement results, conducts physician review and quality control on the analysis results, determines the accuracy and reliability of the analysis, and outputs the comprehensive screening analysis results;

[0119] Based on the feature enhancement results, the comprehensive output submodule initiates a comprehensive image and data analysis process, focusing on the comprehensive analysis of the enhanced image data and related diagnostic data. During this process, each image and data point is thoroughly evaluated, and the analysis results are reviewed and quality controlled by physicians to verify the accuracy and reliability of each analysis. The physician team uses their professional knowledge to review the image diagnosis results to confirm their medical validity, and finally outputs the comprehensive screening analysis results, which records in detail the key information extracted from various modal images and diagnostic data, providing sufficient protection for breast cancer screening.

[0120] See also Figure 7 , a multimodal image fusion breast cancer screening and analysis method, comprising the following steps:

[0121] S1: According to the needs of breast cancer screening, retrieve and collect X-ray, MRI and ultrasound multimodal image data, calculate the image information entropy, and quantitatively evaluate the information contribution of the modality to obtain the information entropy evaluation result;

[0122] S2: Based on the information entropy evaluation results, the weight of the modal image data is dynamically adjusted, and a fast Fourier transform is performed to analyze the frequency points to extract the amplitude and phase information, calculate the energy distribution of each frequency point, and obtain the frequency domain signal characteristics;

[0123] S3: Use frequency domain signal features to perform multimodal frequency domain filtering, synchronously adjust the feature differences between the modalities, apply inverse Fourier transform, reconstruct the time domain image data, and generate frequency domain feature synchronization results;

[0124] S4: Based on the frequency domain feature synchronization results, perform multi-scale decomposition of the image, analyze the feature strength of each scale layer, determine the feature matching points to perform spatial alignment, perform quality analysis on the fused image, and generate fusion quality assessment results;

[0125] S5: Receive the latest imaging data and physiological measurement data, combine the fusion quality assessment results, analyze the impact of the new data on the existing modality weights, update and adjust the weights, reconfigure the image processing process, and generate real-time update results;

[0126] S6: Based on real-time update results, integrate multimodal diagnostic data, unify data formats and timestamps, enhance diagnostic features, conduct physician review and quality control on analysis results, judge analysis accuracy and reliability, and output comprehensive screening analysis results.

[0127] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A multimodal image fusion breast cancer screening and analysis system, characterized by: The system comprises: The information entropy weight adjustment module collects X-ray, MRI, and ultrasound image data, calculates the information entropy of all modal image data, evaluates the information contribution of each modality in the current breast cancer screening scenario, dynamically adjusts the weight according to the information gain comparison results, and generates information entropy weight results; The time-frequency domain feature processing module performs fast Fourier transform on the information entropy weight result, analyzes the features in the time-frequency domain, synchronizes the difference modal features through frequency domain filtering, fuses the frequency domain features, and restores the data to the time domain through inverse Fourier transform to generate frequency domain feature synchronization results; The geometric correction fusion module performs multi-scale decomposition and feature matching on the images of the difference modes based on the frequency domain feature synchronization result, uses spatial correction to adjust the image alignment and performs fusion processing, performs quality assessment, and generates a fusion quality assessment result; The data real-time update module collects diagnostic data in real time, updates the modality weight and correction strategy according to the newly collected diagnostic data and the fusion quality assessment results, adjusts the image processing flow to match the actual conditions of the patients with differences, and generates real-time update results; The comprehensive analysis output module summarizes the real-time update results, performs detail expression optimization, analyzes the image data of each modality, identifies and enhances key diagnostic features, and outputs comprehensive screening analysis results.

2. The multimodal image fusion breast cancer screening and analysis system according to claim 1, characterized in that: The information entropy weight results include weight parameters, entropy values ​​and information gain analysis results, the frequency domain feature synchronization results include frequency domain graphs, time recovery data and filtering features, the fusion quality assessment results include correction parameters, fusion metrics and alignment accuracy, the real-time update results include weight adjustment records, processing parameters and data update logs, and the comprehensive screening analysis results include image quality scores, feature recognition records and data synthesis records.

3. The multimodal image fusion breast cancer screening and analysis system according to claim 1, characterized in that: The information entropy weight adjustment module includes a data acquisition submodule, an information entropy calculation submodule, and a weight adjustment submodule. The data acquisition submodule retrieves and collects X-ray, MRI, and ultrasound multimodal image data from the medical image database according to the needs of breast cancer screening, performs standardized processing on statistical data quality and format, and obtains a multimodal image dataset; The information entropy calculation submodule performs weighted calculation on the information entropy of X-ray, MRI and ultrasound images through the multimodal image data set, quantitatively evaluates the information contribution of the modalities, and obtains the information entropy evaluation result; The weight adjustment submodule dynamically adjusts the weight of the modal image data based on the information entropy evaluation result, performs information gain evaluation and reallocates weight parameters to form an information entropy weight result.

4. The multimodal image fusion breast cancer screening and analysis system according to claim 3, characterized in that: The information entropy is weighted and calculated according to the formula: Calculate the information contribution E of the modal image, where H(X i ) represents the information entropy of the i-th modality image, and M represents the number of difference modalities.

5. The multimodal image fusion breast cancer screening and analysis system according to claim 1, characterized in that: The time-frequency domain feature processing module includes a fast Fourier transform submodule, a frequency domain feature fusion submodule, and an inverse Fourier transform submodule. The fast Fourier transform submodule performs fast Fourier transform based on the information entropy weight result, converts the time series data into frequency domain data, analyzes the frequency points to extract amplitude and phase information, calculates the energy distribution of each frequency point, and obtains the frequency domain signal characteristics; The frequency domain feature fusion submodule adopts the frequency domain signal features to perform multimodal frequency domain filtering, comprehensively processes the frequency domain information of X-ray, MRI and ultrasound image data, synchronously adjusts the feature differences between the modalities, confirms the signal synchronization, and obtains synchronous frequency domain features; The inverse Fourier transform submodule applies inverse Fourier transform through the synchronized frequency domain features, restores the frequency domain feature data to the time domain format, verifies the integrity and practicality of the data, reconstructs the time domain image data, and generates the frequency domain feature synchronization results.

6. The multimodal image fusion breast cancer screening and analysis system according to claim 5, characterized in that: The energy distribution of each frequency point is according to the formula: Calculation is performed, where P(f) represents the energy distribution value of each frequency point, N represents the total number of samples of the time series data, X(n) represents the signal amplitude of the nth point in the time series, the original data is provided directly, and f represents the target frequency. is the adjustment factor.

7. The multimodal image fusion breast cancer screening and analysis system according to claim 1, characterized in that: The geometric correction fusion module includes a multi-scale decomposition submodule, a feature matching submodule, and a fusion quality assessment submodule. The multi-scale decomposition submodule performs multi-scale decomposition of the image based on the frequency domain feature synchronization result, separates the feature layers of difference scale and direction for the image data of the difference mode, analyzes the feature intensity of each scale layer, extracts key edge and texture information, and obtains the decomposed feature layer; The feature matching submodule adopts the decomposed feature layer, determines the feature matching points through pattern recognition, performs spatial alignment, corrects the displacement and rotation differences between images, optimizes the matching accuracy between images, and obtains the corrected alignment result; The fusion quality assessment submodule integrates the difference modality images through the correction alignment results, performs quality analysis on the fused images, evaluates the clarity, contrast and detail retention, and generates a fusion quality assessment result.

8. The multimodal image fusion breast cancer screening and analysis system according to claim 1, characterized in that: The real-time data update module includes a data collection submodule, a weight update submodule, and a process adjustment submodule. The data collection submodule directly receives the latest imaging data and physiological measurement data from the diagnostic equipment through real-time monitoring, performs continuous stream input of data, performs comprehensive information update, and obtains the latest diagnostic data set; The weight update submodule analyzes the impact of new data on existing modality weights through the latest diagnostic data set and the fusion quality assessment results, uses analytical statistics to update and adjust weights, reveals the actual requirements of the latest diagnostic information for image processing, and outputs updated weight parameters; The process adjustment submodule reconfigures the image processing process based on the updated weight parameters, performs process adjustment optimization on each modality image, and generates real-time updated results.

9. The multimodal image fusion breast cancer screening and analysis system according to claim 1, characterized in that: The comprehensive analysis output module includes a result summary submodule, a feature enhancement submodule, and a comprehensive output submodule. The result summary submodule integrates the multimodal diagnostic data through data synchronization based on the real-time update results, unifies the data format and timestamp, checks the consistency and integrity of the data, and generates a standardized diagnostic data set; The feature enhancement submodule enhances the diagnostic features through the standardized diagnostic data set, applies local contrast enhancement and edge sharpening to key areas of the image, enhances the diagnostic information in the image, and outputs the feature enhancement result; The comprehensive output submodule comprehensively analyzes all images and data based on the feature enhancement results, performs physician review and quality control on the analysis results, determines the accuracy and reliability of the analysis, and outputs the comprehensive screening analysis results.

10. A multimodal image fusion breast cancer screening and analysis method, characterized in that: The breast cancer screening and analysis system for multimodal image fusion according to any one of claims 1 to 9 comprises the following steps: According to the needs of breast cancer screening, retrieve and collect X-ray, MRI and ultrasound multimodal image data, calculate the image information entropy, and quantitatively evaluate the information contribution of the modality to obtain the information entropy evaluation result; Based on the information entropy evaluation result, the weight of the modal image data is dynamically adjusted, a fast Fourier transform is performed, the frequency points are analyzed to extract amplitude and phase information, the energy distribution of each frequency point is calculated, and the frequency domain signal characteristics are obtained; Using the frequency domain signal features, multi-modal frequency domain filtering is performed, the feature differences between the modes are synchronously adjusted, an inverse Fourier transform is applied, the time domain image data is reconstructed, and a frequency domain feature synchronization result is generated; Based on the frequency domain feature synchronization result, perform multi-scale decomposition of the image, analyze the feature strength of each scale layer, determine the feature matching points to perform spatial alignment, perform quality analysis on the fused image, and generate a fusion quality assessment result; Receive the latest imaging data and physiological measurement data, combine the fusion quality assessment results, analyze the impact of the new data on the existing modality weights, update and adjust the weights, reconfigure the image processing process, and generate real-time update results; Based on the real-time update results, multimodal diagnostic data are integrated, data formats and timestamps are unified, diagnostic features are enhanced, analysis results are reviewed and quality controlled by physicians, analysis accuracy and reliability are determined, and comprehensive screening analysis results are output.

Images