Bandwidth optimization transmission method based on DICOM image information

Abstract

The invention discloses a bandwidth optimization transmission method based on DICOM image information, and relates to the technical field of medical image information processing, and the method comprises the steps: employing a deep learning model to automatically recognize a key diagnosis information region and a non-key part in a DICOM image; according to an identification result, performing adaptive compression processing on the DICOM file to form an adjusted DICOM file; the multi-dimensional data characteristics in the DICOM file are deeply analyzed, and the multi-dimensional data compression algorithm is adopted to deeply optimize the characteristics of each dimension; the optimized DICOM file is uploaded to a cloud end for distributed storage; based on files stored in a cloud distributed mode, key DICOM file partitions requested by a user are rapidly positioned through an intelligent retrieval framework and a Bloom filter, and the transmission process is optimized. According to the method, the key region of the DICOM image is automatically identified through deep learning, adaptive compression is executed, the transmission efficiency is remarkably improved, key information is ensured to be complete, and the medical service quality is improved.

Description

Technical Field

[0001] The invention relates to the technical field of medical image information processing, in particular to a bandwidth optimization transmission method based on DICOM image information. Background Art

[0002] With the rapid development of medical imaging technology, the DICOM (Digital Imaging and Communications in Medicine) standard has become the main international standard for the exchange, storage, printing and transmission of medical images. The DICOM standard not only specifies the format of image information, but also covers image management and network communication protocols. However, in practical applications, DICOM image files usually contain a large amount of metadata and high-resolution image data, which results in large file sizes and low transmission efficiency, especially when network bandwidth is limited. This has become a problem that needs to be solved urgently.

[0003] Traditional DICOM image transmission methods often use simple compression methods, such as JPEG compression, which can reduce file size to a certain extent, but ignore the difference between key diagnostic information and non-key information in image data, resulting in over-compression of key information, affecting subsequent diagnostic accuracy. In addition, the existing DICOM image file transmission technology lacks effective data management and retrieval mechanisms when processing large-scale data, making it difficult to achieve efficient data access and transmission. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a bandwidth optimization transmission method based on DICOM image information to solve the problems in traditional methods that key diagnostic information may be damaged due to excessive compression and large-scale data management and efficient retrieval mechanisms are missing.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a bandwidth optimization transmission method based on DICOM image information, which includes applying a deep learning model to automatically identify key diagnostic information areas and non-key parts in DICOM images; based on the recognition results, performing adaptive compression processing on the DICOM file to form an adjusted DICOM file; deeply analyzing the multi-dimensional data characteristics in the DICOM file, and using a multi-dimensional data compression algorithm to deeply optimize the characteristics of each dimension; uploading the optimized DICOM file to the cloud for distributed storage; based on the files in the distributed storage on the cloud, quickly locating the key DICOM file partitions requested by the user and optimizing the transmission process through an intelligent retrieval framework and Bloom filters.

[0008] As a preferred solution of the bandwidth optimization transmission method based on DICOM image information of the present invention, wherein: the application of the deep learning model to automatically identify the key diagnostic information area and non-key part in the DICOM image, the specific steps are:

[0009] Use Python's pydicom library to read raw DICOM files, extract image data and perform normalization;

[0010] Based on the normalized image data, histogram equalization and contrast adjustment methods are applied to optimize image quality, and the labelme tool is used to manually annotate key areas.

[0011] The U-Net architecture is selected as the basic model for deep learning and an attention mechanism is added to optimize the deep learning model's ability to focus on key areas;

[0012] For the optimized deep learning model, Dice Loss is combined with the cross entropy loss function to accurately distinguish between key and non-key areas;

[0013] Use the preprocessed image dataset to train the deep learning model;

[0014] Set a threshold based on the probability map output by the trained model to segment the image into key areas and non-key areas;

[0015] Morphological operations are applied to remove noise and smooth boundaries in the critical and non-critical regions, outputting a segmented image in which the critical regions are marked with specific colors.

[0016] As a preferred solution of the bandwidth optimization transmission method based on DICOM image information of the present invention, wherein: the adaptive compression processing is performed on the DICOM file according to the recognition result to form an adjusted DICOM file, and the specific steps are:

[0017] Based on the obtained segmented image results, edge enhancement processing is performed on key areas and non-key areas;

[0018] High resolution is maintained for key areas marked with specific colors, while non-key areas have their resolution appropriately reduced;

[0019] Introducing an adaptive compression ratio determination function C r , calculate the compression ratio of the key area and the non-key area, the expression is:

[0020]

[0021] Among them, C r (pi ,μ p ,σ p ) represents the adaptive compression ratio of the key area and the non-key area, p i is the probability value of the i-th pixel in the key area, μ p is the average probability of the key area, σ p is the standard deviation of the probability of the key area, α is a positive real number that adjusts the steepness of the compression ratio function, β is a positive real number that adjusts the nonlinearity of the compression ratio function, and e is the base of the natural logarithm;

[0022] According to C r The calculated result is that lossless compression is used for the key area, and lossy compression with a high compression ratio is used for the non-key area. The expression is:

[0023] Q f =1-C r (p i ,μ p ,σ p );

[0024] Among them, Q f is the quality factor;

[0025] When C r =1, Q f When it is equal to 0, lossless compression is used in the key area;

[0026] When C r =0, Q f When it is equal to 1, non-critical areas are compressed with high lossy compression ratio and encoded using JPEG 2000 format. The compression ratio is based on Q f Adjustment;

[0027] Use the pydicom library to read the original DICOM file and replace the image data fields of the key and non-key areas that have been segmented and compressed with the compressed image data.

[0028] Save the new DICOM file to form the adjusted DICOM file.

[0029] As a preferred solution of the bandwidth optimization transmission method based on DICOM image information of the present invention, wherein: based on the acquired segmented image results, edge enhancement processing is performed on the key area and the non-key area, and the specific steps are:

[0030] Read the segmented DICOM file through the pydicom library to extract the key and non-key areas in the image data;

[0031] Use the Sobel operator to detect edges in key areas;

[0032] Use Gaussian blur to reduce details in non-critical areas;

[0033] The edge-enhanced critical area is merged with the smoothed non-critical area to form image coherence.

[0034] As a preferred solution of the bandwidth optimization transmission method based on DICOM image information of the present invention, wherein: the multi-dimensional data characteristics of the DICOM file are deeply analyzed, and the multi-dimensional data compression algorithm is used to deeply optimize the characteristics of each dimension. The specific steps are:

[0035] Use the pydicom library to read DICOM files and extract metadata and image data;

[0036] Perform histogram equalization and contrast adjustment on image data to identify key information in metadata;

[0037] Classify metadata into key, secondary and redundant data, use Huffman coding for key metadata, use RLE coding for secondary metadata, and remove duplicate data for redundant metadata;

[0038] Use the pydicom library to read the original DICOM file, fill in the compressed image data and metadata, and save the new DICOM file.

[0039] As a preferred solution of the bandwidth optimization transmission method based on DICOM image information of the present invention, wherein: uploading the optimized DICOM file to the cloud for distributed storage, the specific steps are:

[0040] Based on performance and cost considerations, Alibaba Cloud OSS was selected as the distributed storage service;

[0041] The DICOM file is fragmented using the hash function H, which is defined as

[0042]

[0043] Where f is the file identifier, n is the number of storage nodes, m is the length of the file identifier, i is the index of the number of bits in the first half of the file identifier, j is the index of the number of bits in the second half of the file identifier, and p is the length of the file identifier. i is the ASCII code value of the i-th character in the file identifier, q j is the ASCII code value of the jth character in the file identifier, w i is the weight coefficient of the i-th character in the file identifier, v jis the weight coefficient of the jth character in the file identifier, and mod represents the modulus operator;

[0044] If H(f,n)=0, it means that the file will be assigned to the storage node numbered 0;

[0045] If H(f,n)=1, it means that the file will be assigned to the storage node numbered 1;

[0046] If H(f,n)=2,3,...,n-1, it means that the file will be assigned to the storage nodes numbered 2,3,...,n-1;

[0047] Each file is divided into several data blocks and redundant blocks. The total number of data blocks is composed of the number of original data blocks k plus the number of redundant blocks r, that is, the total number of blocks N is k+r;

[0048] Use Reed-Solomon erasure code to generate data blocks and redundant blocks;

[0049] The optimized DICOM file is sharded using the hash function H, and the redundant blocks generated by Reed-Solomon erasure code are applied to the sharded file data blocks, and the data blocks and redundant blocks are uploaded to the corresponding storage nodes.

[0050] As a preferred solution of the bandwidth optimization transmission method based on DICOM image information of the present invention, the specific steps of quickly locating the key DICOM file partitions requested by the user and optimizing the transmission process are as follows:

[0051] Use Elasticsearch as a full-text search engine to index and search optimized DICOM files;

[0052] Define index mappings to create indexes for each uploaded file, including key metadata and file location information;

[0053] Develop RESTful API to receive user query requests, parse user query conditions, and sort the results according to relevance scores;

[0054] Calculate the bit array size and number of hash functions required for the Bloom filter based on the expected maximum number of files and the desired false positive rate;

[0055] Whenever a new file is uploaded, the hash value of the file is calculated using the hash function in the Bloom filter and the corresponding bit in the bit array is marked;

[0056] According to the mapping relationship between the file identifier and the storage node, a mapping table from the file identifier to the storage node is created;

[0057] When a user request is received, the file identifier and metadata are obtained through the intelligent retrieval framework, and the Bloom filter is used to quickly determine whether the requested file exists in the system;

[0058] If the Bloom filter indicates that the file exists, the storage node where the file is located is determined through the mapping table. If it does not exist, a message indicating that the file is not found is returned;

[0059] Quickly locate and optimize the file transfer process based on the location information of the storage node where the file is located.

[0060] As a preferred solution of the bandwidth optimization transmission method based on DICOM image information of the present invention, the specific steps are as follows: the RESTful API is developed to receive the user's query request, parse the user's query conditions, and sort the results according to the relevance score.

[0061] Define the URL path in the API and accept HTTP GET request parameters including query, fields, and sort request parameters;

[0062] Parsing user-supplied query strings and search fields;

[0063] According to the parsed query conditions, construct a query statement that conforms to the Elasticsearch syntax;

[0064] Use the constructed query statement to call the Elasticsearch query interface to execute the query and obtain the search results returned by Elasticsearch;

[0065] Sort the search results according to the sorting method specified by the user;

[0066] The sorted search results are returned to the user in JSON format.

[0067] In a second aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the bandwidth optimization transmission method based on DICOM image information as described in the first aspect of the present invention is implemented.

[0068] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the bandwidth optimization transmission method based on DICOM image information as described in the first aspect of the present invention is implemented.

[0069] The beneficial effects of the present invention are as follows: the present invention utilizes deep learning technology to automatically identify key diagnostic information areas in images, and accordingly performs adaptive compression processing, thereby significantly improving the transmission efficiency of DICOM image data while ensuring that the integrity and accuracy of key diagnostic information are not compromised; by deeply optimizing DICOM files, including but not limited to key area identification, adaptive compression strategies, the application of multi-dimensional data compression technology, distributed storage mechanisms, and the integration of intelligent retrieval systems, the problems of low transmission efficiency and inconvenient storage management are effectively solved; the present invention is applicable to fields such as medical imaging data centers and telemedicine consultation systems, and can greatly improve the transmission efficiency and storage management capabilities of DICOM image data, providing strong support for the efficient operation of the medical industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0071] Figure 1 This is a flow chart of the bandwidth optimization transmission method based on DICOM image information in Example 1.

[0072] Figure 2 This is a flow chart for deeply optimizing the characteristics of each dimension in Example 1. DETAILED DESCRIPTION

[0073] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0074] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0075] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0076] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, provides a bandwidth optimization transmission method based on DICOM image information, comprising the following steps:

[0077] S1. Apply deep learning models to automatically identify key diagnostic information areas and non-critical parts in DICOM images.

[0078] Furthermore, Python's pydicom library is used to read the original DICOM file, extract the image data and perform normalization;

[0079] Based on the normalized image data, histogram equalization and contrast adjustment methods are applied to optimize image quality, and the labelme tool is used to manually annotate key areas.

[0080] The U-Net architecture is selected as the basic model for deep learning and an attention mechanism is added to optimize the deep learning model's ability to focus on key areas;

[0081] For the optimized deep learning model, Dice Loss is combined with the cross entropy loss function to accurately distinguish between key and non-key areas;

[0082] It should be noted that "for the optimized deep learning model, Dice Loss is combined with the cross entropy loss function to accurately distinguish between key and non-key areas." This step is performed during the model training process. By using Dice Loss combined with the cross entropy loss function, the model can learn how to accurately distinguish between key and non-key areas during the training process.

[0083] Use the preprocessed image dataset to train the deep learning model;

[0084] Set a threshold based on the probability map output by the trained model to segment the image into key areas and non-key areas;

[0085] It should be noted that the step of "setting a threshold based on the probability map output by the trained model to segment the image into key areas and non-key areas" is performed during the model deployment or application phase. The probability map output by the model represents the probability that each pixel belongs to the key area. By setting a threshold, the probability map can be converted into a binary image, thereby segmenting the image into key areas and non-key areas.

[0086] It should also be noted that the specific steps for setting the threshold and segmenting the image into key areas and non-key areas are as follows:

[0087] Use the trained deep learning model to predict the input image and get the probability that each pixel belongs to the key area;

[0088] Draw a histogram of the probability values ​​to observe the distribution characteristics of the probability values ​​in the key and non-key areas. Although this step is not necessary, it can help you understand the data distribution;

[0089] A threshold is automatically selected using the maximum inter-class variance method (Otsu's method). This method can find the threshold that maximizes the variance between two categories (key areas and non-key areas), including calculating all possible thresholds; for each possible threshold, calculating the intra-class variance of the key areas and non-key areas; and selecting the threshold that maximizes the inter-class variance.

[0090] According to the selected threshold, the probability map is converted into a binary image. If the probability of a certain pixel is greater than or equal to the threshold, it is marked as a critical area (for example, assigned a value of 1); if the probability of a certain pixel is less than the threshold, it is marked as a non-critical area (for example, assigned a value of 0).

[0091] Morphological operations are applied to remove noise and smooth boundaries in the critical and non-critical regions, outputting a segmented image in which the critical regions are marked with specific colors.

[0092] It should be noted that "key areas are marked with specific colors" means that during the segmentation process, key areas are usually marked for ease of subsequent processing and identification. The "specific color" here means that in the segmented image, key areas are assigned one or several specific colors to distinguish them from non-key areas. This color marking is added manually or automatically during the segmentation stage to highlight key areas; for example, when manually annotating using the labelme tool, key areas may be assigned one or several specific colors to make it easier to identify these areas in subsequent processing. In the case of automatic segmentation, the segmentation model also outputs a mask image in which key areas are represented by a specific color (usually white or some bright color) and non-key areas are represented by another color (usually black).

[0093] S2. According to the recognition result, adaptive compression processing is performed on the DICOM file to form an adjusted DICOM file.

[0094] Furthermore, based on the obtained segmentation results, edge enhancement processing is performed on the key area and the non-key area;

[0095] It should be noted that based on the obtained segmented image results, edge enhancement processing is performed on the key area and non-key area. The specific steps are as follows:

[0096] Read the segmented DICOM file through the pydicom library to extract the key and non-key areas in the image data;

[0097] Use the Sobel operator to detect the edge of the key area to enhance the boundary of the key area. The Sobel operator highlights the edge information by performing gradient operations in the horizontal and vertical directions of the image.

[0098] Use Gaussian blur to reduce the details of non-critical areas to reduce unnecessary details, which helps to reduce the amount of data in the subsequent compression process. Gaussian blur can smooth the image through convolution operation and reduce the details of non-critical areas;

[0099] Merge the edge-enhanced key area with the smoothed non-key area to form image coherence;

[0100] It should be noted that if necessary, the degree of edge enhancement and smoothing can be fine-tuned according to visual effects.

[0101] High resolution is maintained for key areas marked with specific colors, while non-key areas have their resolution appropriately reduced;

[0102] Introducing an adaptive compression ratio determination function C r , determine the compression ratio of the key area and the non-key area, the expression is:

[0103]

[0104] Among them, C r (p i ,μ p ,σ p ) represents the adaptive compression ratio of the key area and the non-key area, p i is the probability value of the i-th pixel in the key area, p i is the probability value of the i-th pixel in the key area, μ p is the average probability of the key area, σ p is the standard deviation of the probability of the key area, α is a positive real number that adjusts the steepness of the compression ratio function, β is a positive real number that adjusts the nonlinearity of the compression ratio function, e is the base of the natural logarithm, Represents the denominator of a Sigmoid function, which is used to map the probability value to a value in the interval [0,1], reflecting the importance of the key area. Represents an adjusted Sigmoid function, which is based on the pixel probability value p in the key area i To determine the compression ratio, |p i -μ p | represents the absolute value symbol represents the absolute value of the difference between the probability value of the i-th pixel in the key region and the average probability of the key region, Represents a ratio related to the standard deviation and the difference in probability values, which is used to smooth the change in compression ratio and ensure that the importance of key areas is not overly magnified or reduced due to extremely small probability differences. Represents a value in the range [0,1], which is used to further adjust the compression ratio to ensure that the compression ratio of the key area is not too high;

[0105] Preferably, an adaptive compression ratio determination function C is introduced r The advantages are as follows:

[0106] Achieving precise compression: Through C r The function can automatically adjust the compression ratio according to the importance of key areas and non-key areas, thereby effectively reducing the storage space occupied by non-key information while ensuring the clarity of key information.

[0107] Outstanding flexibility: C r The function can flexibly adjust the compression ratio according to the probability distribution of the key area, so that the key area maintains high quality, while the non-key area can accept a higher compression ratio to adapt to different compression requirements.

[0108] Efficient storage: By adaptively compressing non-critical areas, the overall size of the DICOM file can be significantly reduced without affecting key diagnostic information, thereby improving storage efficiency and transmission speed.

[0109] Personalized treatment: C r The function can perform personalized compression processing based on the specific content of the image, ensuring that the importance of each pixel is fully considered, thereby achieving the best compression effect.

[0110] Parameter adjustment: By adjusting the parameters α and β, the rate of change and degree of nonlinearity of the compression ratio can be flexibly controlled to adapt to different application scenarios.

[0111] Easy to integrate: Due to C r The design of the functions is based on existing techniques (such as probability maps output by deep learning models), so they can be easily integrated into existing DICOM image processing pipelines.

[0112] According to C r The calculated result is that lossless compression is used for the key area, and lossy compression with a high compression ratio is used for the non-key area. The expression is:

[0113] Q f =1-C r (p i ,μ p ,σ p );

[0114] Among them, Q fis the quality factor, which is used to determine the degree of image compression;

[0115] When C r When it is equal to 1 (indicating a critical area), Q f When it is equal to 0, the key area adopts lossless compression, which means that the image data in the key area can be completely restored to its original state after compression without any information loss;

[0116] When C r When it is equal to 0 (indicating a non-critical area), Q f =1, non-critical areas use lossy compression with high compression ratio, and are encoded using JPEG 2000 format. The compression ratio is based on Q f Adjustment, which means that the image data in non-critical areas will lose some information during the compression process, but it also means that a higher compression rate can be achieved, thus saving storage space;

[0117] In short, the quality factor Q f With adaptive compression ratio C r Inversely proportional relationship.

[0118] Use the pydicom library to read the original DICOM file and replace the image data field with the compressed image data;

[0119] Save the new DICOM file, ensuring that all metadata remain intact, to form the adjusted DICOM file.

[0120] It should be noted that C r The value range of Q is [0,1], where a value close to 0 indicates a higher compression ratio, which is suitable for non-critical areas; a value close to 1 indicates a lower compression ratio or lossless compression, which is suitable for critical areas; f The value range of is also [0,1]. A value close to 0 indicates a higher degree of compression (ie, close to lossless), and a value close to 1 indicates a lower degree of compression or a higher compression ratio.

[0121] S3. In-depth analysis of the multi-dimensional data characteristics in DICOM files, and use multi-dimensional data compression algorithms to deeply optimize the characteristics of each dimension.

[0122] Furthermore, the pydicom library is used to read DICOM files and extract metadata and image data;

[0123] It should be noted that the specific implementation steps of "extracting metadata and image data" are: use the pydicom.dcmread function to read the file and extract the PixelData field and other metadata fields respectively;

[0124] Perform histogram equalization and contrast adjustment on image data to identify key information in metadata;

[0125] Among them, "histogram equalization and contrast adjustment of image data" means using cv2.equalizeHist or skimage.exposure.equalize_adapthist to perform histogram equalization; using cv2.convertScaleAbs or skimage.exposure.rescale_intensity to adjust the contrast;

[0126] Among them, "identifying key information in metadata" refers to defining the standards for key metadata, such as imaging parameters, patient information, etc.; using statistical methods (such as mean, standard deviation, etc.) to evaluate the range and distribution of each metadata item;

[0127] Classify metadata into key, secondary and redundant data, use Huffman coding for key metadata, use RLE coding for secondary metadata to reduce repeated sequences, and remove duplicate data for redundant metadata;

[0128] It should be noted that "redundant metadata deduplication" means identifying and removing duplicate data by comparing adjacent metadata items.

[0129] For example, imaging parameters and patient identifiers are considered key metadata; date and time stamps, etc. are considered secondary metadata; and duplicate patient IDs, etc. are considered redundant metadata.

[0130] Use the pydicom library to read the original DICOM file, fill in the compressed image data and metadata, and save the new DICOM file.

[0131] It should be noted that after "saving a new DICOM file", a testing and verification phase is required; compare the DICOM files before and after compression to evaluate the compression effect. The main implementation steps are as follows: compare the size of DICOM files before and after compression; perform image quality assessment to ensure that the image quality of key areas is not damaged; and verify the integrity and correctness of the metadata.

[0132] It should also be noted that the testing and verification phase is very important because it ensures the effectiveness and reliability of the entire technical solution. Its main functions include the following aspects:

[0133] By comparing the size of DICOM files before and after compression, the effect of the compression algorithm, i.e. the compression ratio, can be intuitively evaluated to ensure that the compressed image still meets the needs of diagnosis or analysis. This usually involves visual inspection and measurement of quantitative indicators (such as PSNR, SSIM, etc.) to ensure that the image quality in key areas has not been significantly reduced.

[0134] Check whether any metadata is lost or modified by mistake during the compression process, and ensure that all key and secondary metadata are correctly retained and encoded; ensure that the metadata in the compressed new DICOM file is completely consistent with the original file, especially those metadata that are critical for diagnosis and clinical decision-making.

[0135] Through actual testing, the overall performance of the solution is evaluated, including the comprehensive performance in terms of compression efficiency, image quality and metadata processing; potential technical problems or defects are discovered and corrected in a timely manner to improve the stability and practicality of the solution.

[0136] Ensure that the final DICOM file complies with the DICOM standard and relevant legal and regulatory requirements for medical image processing; ensure that all changes are recorded for easy audit and quality control.

[0137] S4. Upload the optimized DICOM file to the cloud for distributed storage.

[0138] Furthermore, based on performance and cost considerations, Alibaba Cloud OSS is selected as the distributed storage service; the hash function H is used to fragment the DICOM file, and the hash function H is defined as

[0139]

[0140] Where f is the file identifier, n is the number of storage nodes, m is the length of the file identifier, i is the index of the number of bits in the first half of the file identifier, j is the index of the number of bits in the second half of the file identifier, and p is the length of the file identifier. i is the ASCII code value of the i-th character in the file identifier, q j is the ASCII code value of the jth character in the file identifier, w i is the weight coefficient of the i-th character in the file identifier, v j is the weight coefficient of the jth character in the file identifier, and mod represents the modulus operator;

[0141] It should be noted that the value range of i is from 1 to Where m is the length of the file identifier. Therefore, i represents the position of a certain bit in the first half of the file identifier; j ranges from To m. Therefore, j represents the position of a certain bit in the second half of the file identifier.

[0142] Preferably, the modulus operation mod ensures that the range of H(f,n) is [0,n-1], which means that the result of the hash function will always be between 0 and n-1 (including the two end points), which corresponds exactly to the numbering range of the storage nodes.

[0143] If H(f,n)=0, it means that the file will be assigned to the storage node numbered 0;

[0144] If H(f,n)=1, it means that the file will be assigned to the storage node numbered 1;

[0145] If H(f,n)=2,3,...,n-1, it means that the file will be assigned to the storage nodes numbered 2,3,...,n-1;

[0146] Each file is divided into several data blocks and redundant blocks. The total number of data blocks is composed of the number of original data blocks k plus the number of redundant blocks r, that is, the total number of blocks N is k+r;

[0147] Use Reed-Solomon erasure code to generate data blocks and redundant blocks;

[0148] Use the hash function H to shard the optimized DICOM file, apply the Reed-Solomon erasure code to the sharded file data blocks to generate redundant blocks, and upload the data blocks and redundant blocks to the corresponding storage nodes;

[0149] Preferably, Reed-Solomon erasure code is an existing mature technology, which is widely used in the fields of data storage and communication. It is mainly used to improve the reliability and fault tolerance of data, especially in distributed storage systems. Reed-Solomon erasure code can ensure that even if part of the data is lost or damaged, the original data can be restored through the remaining data and redundant information.

[0150] S5. Based on the files stored in the cloud distributedly, the intelligent retrieval framework and Bloom filter are used to quickly locate the key DICOM file partitions requested by the user and optimize the transmission process.

[0151] Furthermore, we use Elasticsearch as a full-text search engine to index and search the optimized DICOM files;

[0152] Define index mapping to create an index for each uploaded file, including key metadata (such as patient ID, study date, image type, etc.) and file location information, laying the foundation for subsequent rapid retrieval;

[0153] Develop RESTful APIs to receive user query requests, parse user query conditions, and sort results based on relevance scores to ensure that users can obtain the most relevant results;

[0154] It should be noted that "develop a RESTful API to receive user query requests, parse the user's query conditions, and sort the results according to the relevance score to ensure that users can obtain the most relevant results." The specific steps are to define the URL path in the API and accept HTTP GET request parameters including query (query string), fields (specify which fields to search) and sort (sort by) request parameters;

[0155] Parsing user-supplied query strings and search fields;

[0156] It should be noted that "parsing the query string and search fields provided by the user" means parsing the query to ensure that it conforms to the query syntax of Elasticsearch, validating the fields parameter to confirm whether the specified fields are legal, and confirming the sort parameter to ensure that the sorting method is valid;

[0157] According to the parsed query conditions, construct a query statement that conforms to the Elasticsearch syntax and ensure that the query statement correctly references the fields specified in the fields parameter;

[0158] Use the constructed query statement to call the Elasticsearch query interface to execute the query and obtain the search results returned by Elasticsearch;

[0159] Sort the search results according to the sorting method specified by the user;

[0160] It should be noted that if the sort parameter is empty, the default sorting method will be used;

[0161] Return the sorted search results to the user in JSON format;

[0162] When returning, you need to ensure that the format of the returned JSON data is correct, including necessary metadata (such as search time, total hit count, etc.).

[0163] The bit array size m and the number of hash functions required by the Bloom filter are calculated based on the expected maximum number of files N and the expected false alarm rate p to efficiently determine whether a file exists;

[0164] It should be noted that the expression used to calculate the bit array size m to ensure that the Bloom filter can work correctly given the maximum number of files N and the expected false alarm rate p is:

[0165]

[0166] It should be noted that the number k used to calculate the hash function to ensure that the Bloom filter can achieve the desired false positive rate under the given bit array size m and the number of files N is expressed as:

[0167]

[0168] Whenever a new file is uploaded, the hash value of the file is calculated using the hash function in the Bloom filter and the corresponding bit is marked in the bit array to ensure that all files are correctly marked;

[0169] Based on the mapping relationship between file identifiers and storage nodes, a mapping table from file identifiers to storage nodes is created, so that a clear connection is established between the file and its storage location;

[0170] When a user request is received, the file identifier and metadata are obtained through the intelligent retrieval framework, and the Bloom filter is used to quickly determine whether the requested file exists in the system, thereby avoiding unnecessary queries;

[0171] If the Bloom filter indicates that the file exists, the mapping table is used to determine the storage node where the file is located. If it does not exist, a message indicating that the file is not found is returned to ensure that users can receive timely feedback.

[0172] Based on the location information of the storage node where the file is located, the file transfer process is quickly located and optimized to achieve efficient data transmission, for example, by using efficient transmission protocols (such as HTTP / 2 or HTTP / 3) and preloading strategies to reduce latency.

[0173] This embodiment also provides a computer device, which is applicable to the case of the bandwidth optimization transmission method based on DICOM image information, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to implement the bandwidth optimization transmission method based on DICOM image information proposed in the above embodiment.

[0174] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0175] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the bandwidth optimization transmission method based on DICOM image information proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0176] In summary, the present invention uses deep learning technology to automatically identify key diagnostic information areas in images, and performs adaptive compression processing accordingly, thereby significantly improving the transmission efficiency of DICOM image data while ensuring that the integrity and accuracy of key diagnostic information are not damaged; by deeply optimizing DICOM files, including but not limited to key area identification, adaptive compression strategy, application of multi-dimensional data compression technology, distributed storage mechanism and integration of intelligent retrieval system, the problems of low transmission efficiency and inconvenient storage management are effectively solved; the present invention is suitable for fields such as medical imaging data centers and telemedicine consultation systems, and can greatly improve the transmission efficiency and storage management capabilities of DICOM image data, providing strong support for the efficient operation of the medical industry.

[0177] Example 2, referring to Table 1, is the second example of the present invention. To further verify the advancement of the present invention, experimental simulation data of a bandwidth optimization transmission method based on DICOM image information is provided.

[0178] Two different groups of DICOM image files were selected: one group was processed using a traditional compression method, and the other group was processed using the method of the present invention. Each group of files contained 100 images, and the image size was 512x512 pixels.

[0179] First, the pydicom library was used to read the original DICOM file and extract the image data for normalization. Histogram equalization and contrast adjustment methods were applied to optimize the image quality, and the labelme tool was used to manually annotate the key areas.

[0180] Secondly, the U-Net architecture is selected as the basic model and the attention mechanism is added to optimize the deep learning model's ability to focus on key areas. Dice Loss is combined with the cross entropy loss function to accurately distinguish key and non-key areas. The deep learning model is trained using the preprocessed image dataset, and the threshold is set according to the probability map output by the trained model to segment the image into key and non-key areas. Morphological operations are applied to remove noise and smooth boundaries in key and non-key areas, and the segmented image is output, in which the key areas are marked with specific colors.

[0181] Next, based on the obtained segmented image results, edge enhancement processing is performed on the key area and non-key area. r The calculated results use lossless compression for key areas and lossy compression with high compression ratio for non-key areas. At the same time, the JPEG 2000 format is used for encoding. The compression ratio is based on Q fAdjust, use pydicom library to read DICOM files, extract metadata and image data. Classify metadata into key, secondary and redundant data, and use Huffman encoding, RLE encoding and deduplication operations respectively.

[0182] Furthermore, Alibaba Cloud OSS is selected as the distributed storage service. The DICOM file is sharded using the hash function H, and the redundant blocks generated by Reed-Solomon erasure code are applied to the sharded file data blocks, and the data blocks and redundant blocks are uploaded to the corresponding storage nodes.

[0183] Finally, we use Elasticsearch as a full-text search engine to index and search the optimized DICOM files, develop a RESTful API to receive user query requests, and sort the results according to the relevance score. We use Bloom filters to quickly determine whether the requested file exists in the system and optimize the file transfer process.

[0184] The details are shown in Table 1 below:

[0185] Table 1 Data table of bandwidth optimization transmission implementation example based on DICOM image information

[0186]

[0187] It can be seen from the above table that the size of the DICOM file processed by the method of the present invention is significantly reduced compared to the original file, and the compression ratio reaches between 20% and 30%. In addition, due to the use of an optimized compression algorithm, the file transmission time is also effectively shortened. For example, the transmission time of DICOM_01 is reduced from the expected 20 seconds to 15 seconds, effectively reducing the occupation of network bandwidth.

[0188] In terms of storage space utilization, a high utilization rate of about 90% is achieved through the distributed storage strategy and the reasonable allocation of redundant data blocks. This means that more DICOM files can be stored under the same storage capacity, which improves the efficiency of storage resource utilization.

[0189] In summary, the present invention automatically identifies the key diagnostic information area and non-key parts in DICOM images through a deep learning model, and performs adaptive compression processing on them, which not only reduces the file size, but also reduces the transmission time and improves the storage space utilization. Compared with traditional compression methods, this method significantly improves the transmission efficiency and storage efficiency of DICOM image information, and has obvious innovation and advantages.

[0190] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A bandwidth optimization transmission method based on DICOM image information, characterized in that: include, Apply deep learning models to automatically identify key diagnostic information areas and non-critical parts in DICOM images; According to the recognition result, the DICOM file is adaptively compressed to form an adjusted DICOM file; In-depth analysis of the multi-dimensional data characteristics of DICOM files, and use of multi-dimensional data compression algorithms to deeply optimize the characteristics of each dimension; Upload the optimized DICOM files to the cloud for distributed storage; Based on the files stored in the cloud distributedly, the intelligent retrieval framework and Bloom filter are used to quickly locate the key DICOM file partitions requested by users and optimize the transmission process.

2. The bandwidth optimization transmission method based on DICOM image information according to claim 1, characterized in that: The application of the deep learning model to automatically identify the key diagnostic information area and non-key parts in the DICOM image is specifically performed in the following steps: Use Python's pydicom library to read raw DICOM files, extract image data and perform normalization; Based on the normalized image data, histogram equalization and contrast adjustment methods are applied to optimize image quality, and the labelme tool is used to manually annotate key areas. The U-Net architecture is selected as the basic model for deep learning and an attention mechanism is added to optimize the deep learning model's ability to focus on key areas; For the optimized deep learning model, Dice Loss is combined with the cross entropy loss function to accurately distinguish between key and non-key areas; Use the preprocessed image dataset to train the deep learning model; Set a threshold based on the probability map output by the trained model to segment the image into key areas and non-key areas; Morphological operations are applied to remove noise and smooth boundaries in the critical and non-critical regions, outputting a segmented image in which the critical regions are marked with specific colors.

3. The bandwidth optimization transmission method based on DICOM image information according to claim 2, characterized in that: According to the recognition result, the DICOM file is adaptively compressed to form an adjusted DICOM file. The specific steps are: Based on the obtained segmented image results, edge enhancement processing is performed on key areas and non-key areas; High resolution is maintained for key areas marked with specific colors, while non-key areas have their resolution appropriately reduced; Introducing an adaptive compression ratio determination function C r , calculate the compression ratio of the key area and the non-key area, the expression is: Among them, C r (p i , μ p , σ p ) represents the adaptive compression ratio of the key area and the non-key area, p i is the probability value of the i-th pixel in the key area, μ p is the average probability of the key area, σ p is the standard deviation of the probability of the key area, α is a positive real number that adjusts the steepness of the compression ratio function, β is a positive real number that adjusts the nonlinearity of the compression ratio function, and e is the base of the natural logarithm; According to C r The calculated result is that lossless compression is used for the key area, and lossy compression with a high compression ratio is used for the non-key area. The expression is: Q i =1-C r (p i ,μ p ,σ p ); Among them, Q f is the quality factor; When C r =1, Q f When it is equal to 0, lossless compression is used in the key area; When C r =0, Q f When it is equal to 1, non-critical areas are compressed with high lossy compression ratio and encoded using JPEG 2000 format. The compression ratio is based on Q f Adjustment; Use the pydicom library to read the original DICOM file and replace the image data fields of the key and non-key areas that have been segmented and compressed with the compressed image data. Save the new DICOM file to form the adjusted DICOM file.

4. The bandwidth optimization transmission method based on DICOM image information according to claim 3, characterized in that: Based on the obtained segmented image results, edge enhancement processing is performed on the key area and the non-key area. The specific steps are: Read the segmented DICOM file through the pydicom library to extract the key and non-key areas in the image data; Use the Sobel operator to detect edges in key areas; Use Gaussian blur to reduce details in non-critical areas; The edge-enhanced critical area is merged with the smoothed non-critical area to form image coherence.

5. The bandwidth optimization transmission method based on DICOM image information according to claim 4, characterized in that: The in-depth analysis of the multi-dimensional data characteristics in the DICOM file and the use of a multi-dimensional data compression algorithm to deeply optimize the characteristics of each dimension are as follows: Use the pydicom library to read DICOM files and extract metadata and image data; Perform histogram equalization and contrast adjustment on image data to identify key information in metadata; Classify metadata into key, secondary and redundant data, use Huffman coding for key metadata, use RLE coding for secondary metadata, and remove duplicate data for redundant metadata; Use the pydicom library to read the original DICOM file, fill in the compressed image data and metadata, and save the new DICOM file.

6. The bandwidth optimization transmission method based on DICOM image information according to claim 5, characterized in that: The optimized DICOM file is uploaded to the cloud for distributed storage. The specific steps are: Based on performance and cost considerations, Alibaba Cloud OSS was selected as the distributed storage service; The DICOM file is fragmented using the hash function H, which is defined as Where f is the file identifier, n is the number of storage nodes, m is the length of the file identifier, i is the index of the number of bits in the first half of the file identifier, jj is the index of the number of bits in the second half of the file identifier, and p f is the ASCII code value of the i-th character in the file identifier, q j is the ASCII code value of the jth character in the file identifier, w i is the weight coefficient of the i-th character in the file identifier, v j is the weight coefficient of the jth character in the file identifier, and mod represents the modulus operator; If H(f, n) = 0, it means that the file will be assigned to the storage node numbered 0; If H(f, n) = 1, it means that the file will be assigned to the storage node numbered 1; If H(f, n) = 2, 3, ..., n-1, it means that the file will be assigned to the storage nodes numbered 2, 3, ..., n-1; Divide each file into several data blocks and redundant blocks. The total number of data blocks is composed of the number of original data blocks kk plus the number of redundant blocks r, that is, the total number of blocks N is kk+r; Use Reed-Solomon erasure code to generate data blocks and redundant blocks; The optimized DICOM file is sharded using the hash function H, and the redundant blocks generated by Reed-Solomon erasure code are applied to the sharded file data blocks, and the data blocks and redundant blocks are uploaded to the corresponding storage nodes.

7. The bandwidth optimization transmission method based on DICOM image information according to claim 6, characterized in that: The specific steps of quickly locating the key DICOM file partitions requested by the user and optimizing the transmission process are as follows: Use Elasticsearch as a full-text search engine to index and search optimized DICOM files; Define index mappings to create indexes for each uploaded file, including key metadata and file location information; Develop RESTful API to receive user query requests, parse user query conditions, and sort the results according to relevance scores; Calculate the bit array size and number of hash functions required for the Bloom filter based on the expected maximum number of files and the desired false positive rate; Whenever a new file is uploaded, the hash value of the file is calculated using the hash function in the Bloom filter and the corresponding bit in the bit array is marked; According to the mapping relationship between the file identifier and the storage node, a mapping table from the file identifier to the storage node is created; When a user request is received, the file identifier and metadata are obtained through the intelligent retrieval framework, and the Bloom filter is used to quickly determine whether the requested file exists in the system; If the Bloom filter indicates that the file exists, the storage node where the file is located is determined through the mapping table. If it does not exist, a message indicating that the file is not found is returned; Quickly locate and optimize the file transfer process based on the location information of the storage node where the file is located.

8. The bandwidth optimization transmission method based on DICOM image information according to claim 7, characterized in that: The RESTful API is developed to receive user query requests, parse user query conditions, and sort the results according to relevance scores. The specific steps are: Define the URL path in the API and accept HTTP GET request parameters including query, fields, and sort request parameters; Parsing user-supplied query strings and search fields; According to the parsed query conditions, construct a query statement that conforms to the Elasticsearch syntax; Use the constructed query statement to call the Elasticsearch query interface to execute the query and obtain the search results returned by Elasticsearch; Sort the search results according to the sorting method specified by the user; The sorted search results are returned to the user in JSON format.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the bandwidth optimization transmission method based on DICOM image information according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the bandwidth optimization transmission method based on DICOM image information according to any one of claims 1 to 8 are implemented.

Images