DICOM file sequence processing method, medical image analysis method, system and device, computer storage medium and computer program product

Abstract

The invention provides a DICOM (Digital Imaging and Communications in Medicine) file sequence processing method, a medical image analysis method, system and device, a computer storage medium and a computer program product, the processing method comprises the following steps: loading a DICOM file sequence corresponding to a series of image sets obtained by tomography, and arranging according to a scanning sequence to obtain an initial image file queue; reading DICOM files from a specified position to form a key image file queue with a fixed length; and reading sequence-level DICOM metadata and image-level DICOM metadata for representing key features from the key image file queue, and arranging the sequence-level DICOM metadata and the image-level DICOM metadata according to a preset rule to obtain a sequence DICOM metadata queue with fixed length and fixed information items of each position. By adopting the scheme provided by the invention, the obtained metadata queue with the fixed dimension provides a data basis for subsequent deep learning network input, and improvement of the adaptability with the deep learning network and improvement of the analysis accuracy are facilitated.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to methods for processing DICOM file sequences, medical image analysis methods, systems, devices, computer storage media, and computer program products. Background Technology

[0002] Currently, commonly used tomographic scanning methods in clinical practice include computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET). A complete examination typically yields multiple scan sequences, each containing a set of images of a specified area of ​​the patient, usually ranging from tens to thousands of consecutive tomographic images. These images are stored in the Digital Imaging and Communications in Medicine (DICOM) standard format. DICOM metadata, in the form of a standardized tag list, records key information such as examination type, scan parameters, scan area, and patient information, providing crucial foundational data for medical image analysis.

[0003] With the exponential growth of medical imaging data, hospitals are generating massive amounts of DICOM files daily. Traditional manual review and analysis methods are not only inefficient and time-consuming, but also fail to meet the timeliness requirements of clinical diagnosis, especially when dealing with complex cases, where doctors often face data backlogs. Deep learning, with its powerful feature extraction and pattern recognition capabilities, offers new breakthroughs for intelligent medical image analysis. Applying it to the processing of DICOM metadata is expected to significantly improve analysis efficiency and play a vital role in areas such as early disease detection and lesion identification. However, current technology applications face significant challenges. Directly merging the DICOM header information of all images within a sequence into a feature vector presents two fundamental technical bottlenecks: First, the problem of data redundancy. Since a single sequence can contain hundreds or even thousands of images, the integrated feature vector contains a large amount of repetitive information, increasing computational burden and potentially interfering with effective feature extraction. Second, the problem of inconsistent data dimensionality. The number of images in different scan sequences varies significantly, resulting in feature vectors of varying lengths. Deep learning networks have extremely high requirements for the consistency of input data dimensionality. This contradiction makes it difficult for existing technologies to achieve efficient model training and clinical applications, becoming a key obstacle restricting the development of intelligent medical image analysis.

[0004] Therefore, how to effectively remove redundant information from DICOM metadata while retaining key diagnostic information, and how to establish a standardized data processing mechanism to unify feature vector length in order to adapt to the requirements of deep learning networks for simplified and fixed-dimensional input data, are urgent technical problems to be solved. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method for processing DICOM file sequences, a method for analyzing medical images, a system, an apparatus, a computer storage medium, and a computer program product, which can efficiently remove redundant metadata and unify feature vector lengths while retaining key diagnostic information, meet the input requirements of deep learning networks, and improve the efficiency and accuracy of intelligent medical image analysis.

[0006] In a first aspect, embodiments of the present invention provide a method for processing DICOM file sequences, comprising: loading DICOM file sequences corresponding to a series of images obtained from computed tomography scans, and arranging them according to the scanning order to obtain an initial image file queue; reading DICOM files from a specified position in the initial image file queue to form a key image file queue of fixed length; reading sequence-level DICOM metadata and image-level DICOM metadata representing key features from the key image file queue, arranging them according to a preset rule to obtain a sequence DICOM metadata queue of fixed length and with fixed information items at each position; the key features include: examination type, scanning parameters, scanning site, and patient information.

[0007] This scheme first arranges DICOM files in scanning order to form an initial image file queue. Then, DICOM files are read from a specified location to form a fixed-length key image file queue. Sequence-level DICOM metafiles and image-level DICOM metafiles are read from the key image file queue to obtain a fixed-length sequence DICOM metadata queue. This effectively eliminates data redundancy in the original sequence, compresses hundreds to thousands of frames of data into a few key frames, significantly reduces the computational burden, and at the same time, selectively preserves core diagnostic information. The resulting fixed-dimensional metadata queue provides a data foundation for subsequent input to deep learning networks, which is beneficial for improving the adaptability to deep learning networks and enhancing the accuracy of analysis.

[0008] In conjunction with the first aspect, in a first possible implementation of the first aspect, reading the DICOM file from a specified position in the initial image file queue includes: sequentially reading the first DICOM file, a DICOM file at an intermediate position, and the last DICOM file; or, sequentially reading the first DICOM file and the last DICOM file.

[0009] The above scheme specifies the designated position in the initial image file queue, which can be the first or last frame, or the first or last frame plus intermediate frames. This strategy reduces data redundancy in the original sequence while preserving key features, which helps to balance diagnostic efficiency and information integrity during analysis.

[0010] In conjunction with the first aspect, in the second possible implementation of the first aspect, obtaining a sequence DICOM metadata queue with a fixed length and fixed information items at each position includes: filling the sequence DICOM metadata queue with tag values ​​from the read DICOM file; when a tag value is missing, filling it with "" if the data type corresponding to the tag value is text; filling it with NaN if the data type corresponding to the tag value is numeric; filling it with "" if the data type corresponding to the tag value is enumeration; and normalizing the numeric value if the data type corresponding to the tag value is a value within a continuous value range.

[0011] The above scheme provides differentiated missing value imputation strategies (e.g., filling in “” for text and NaN for numerical values) based on different data types when constructing a fixed-length sequence DICOM metadata queue, and performs normalization processing on continuous numerical values. This scheme provides a standardized processing method for data consistency, which helps to avoid model training bias caused by missing values ​​and provides a data foundation for adapting to the input requirements of deep learning networks.

[0012] Secondly, embodiments of the present invention provide a medical image analysis method, comprising: constructing a DICOM metadata queue based on medical images using a processing method provided in the first aspect or any possible implementation of the first aspect; inputting the sequence DICOM metadata queue into a deep learning network, the deep learning network comprising a feature preprocessing and classification layer, a feature encoding layer, a feature fusion layer, a context modeling layer, and an output layer; triggering the deep learning network to convert the sequence DICOM metadata queue into a fixed-dimensional feature vector; and triggering the deep learning network to analyze the metadata semantic information contained in the feature vector and the visual features of the medical images.

[0013] Deep learning networks are machine learning models based on multi-layered neural networks. By simulating the hierarchical structure of neurons in the human brain, they automatically extract high-order abstract features from data. Their core features include: multi-layered nonlinear transformations: through structures such as convolutional layers and recurrent layers, they extract features from data layer by layer, making them suitable for processing complex data such as images, speech, and text. They also possess end-to-end learning capabilities: eliminating the need for manual feature design, they directly learn effective representations from raw data, simplifying the feature engineering process in traditional machine learning.

[0014] This approach involves inputting a sequence of DICOM metadata, constructed using methods for processing DICOM file sequences, into a deep learning network. After processing by the feature preprocessing and classification layers, feature encoding layers, feature fusion layers, context modeling layers, and output layers within the deep learning network, a fixed-dimensional feature vector is obtained. This vector is then used to trigger the deep learning network to analyze the metadata semantic information contained in the feature vector and the visual features of the medical image. This achieves joint analysis of metadata semantic information and the visual features of the medical image, which helps improve the accuracy of the analysis and compensates for the low accuracy of pure image analysis due to the lack of clinical background.

[0015] In conjunction with the second aspect, in the first possible implementation of the second aspect, the feature preprocessing and classification layer classifies the input sequence DICOM metadata queue according to data type; including: categorical features are converted by One-hot encoding or embedding technology, continuous features are processed by standardization or Min-Max normalization, text features are processed by sequence encoding method, and missing values ​​are processed by masking.

[0016] The above scheme categorizes feature preprocessing and classification layers according to data type, standardizes the mathematical expression of different types of data, solves the problem of inconsistent formats of multimodal data, and improves the model convergence speed. In addition, the scheme enhances the model's generalization ability in clinical data by masking missing values.

[0017] In conjunction with the second aspect, in a second possible implementation of the second aspect, the feature fusion layer concatenates the preprocessed and encoded features of each data in the sequence DICOM metadata queue, and applies a multi-head self-attention mechanism to dynamically adjust the importance weight of each metadata field.

[0018] The above scheme uses a feature fusion layer to splice preprocessed and encoded features and apply a multi-head self-attention mechanism to dynamically adjust the weights of each metadata field, automatically focus on key diagnostic features such as scanning parameters, and suppress redundant interference such as equipment information. This helps improve the accuracy of medical image analysis. In addition, by capturing the semantic relationships between fields, it helps generate more expressive feature vectors and supports multi-dimensional analysis of complex cases.

[0019] Thirdly, embodiments of the present invention provide a medical image analysis system, comprising: a loading module, a key queue construction module, a metadata queue construction module, and a deep learning network. The loading module is used to load a sequence of DICOM files corresponding to a series of images obtained from tomographic scans, and arrange them according to the scan order to obtain an initial image file queue. The key queue construction module is used to read DICOM files from a specified position in the initial image file queue to form a key image file queue of fixed length. The metadata queue construction module is used to read sequence-level DICOM metadata and image-level DICOM metadata representing key features from the key image file queue, and arrange them according to a preset rule to obtain a sequence DICOM metadata queue of fixed length with fixed information items at each position. The key features include: scan parameters, device information, and patient location. The deep learning network is used to convert the sequence DICOM metadata queue constructed by the metadata queue construction module into a fixed-dimensional feature vector; and to perform analysis based on the metadata semantic information contained in the feature vector and the visual features of the medical images.

[0020] Fourthly, embodiments of the present invention provide a medical image analysis apparatus, the apparatus comprising: a processor configured to execute computer-executable instructions; and a memory storing one or more computer-executable instructions, wherein when executed by the processor, the computer-executable instructions implement the steps of the method for processing a DICOM file sequence as described in the first aspect or any possible implementation thereof, or implement the steps of the medical image analysis method as described in the second aspect or any possible implementation thereof.

[0021] Fifthly, embodiments of the present invention provide a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for processing DICOM file sequences as described in the first aspect or any possible implementation of the first aspect, or the medical image analysis method as described in the second aspect or any possible implementation of the second aspect.

[0022] In a sixth aspect, embodiments of the present invention provide a computer program product, including a computer program that, when executed by a processor, implements a method for processing DICOM file sequences as described in the first aspect or any possible implementation of the first aspect, or a medical image analysis method as described in the second aspect or any possible implementation of the second aspect.

[0023] In a seventh aspect, embodiments of the present invention provide a chip system applied to an electronic device. The chip system includes one or more processors, which are configured to invoke computer instructions to cause the electronic device to perform a method for processing a DICOM file sequence as described in the first aspect or any possible implementation thereof, or a method for analyzing medical images as described in the second aspect or any possible implementation thereof.

[0024] Understandably, the technical effects achieved by the medical image analysis system described in the third aspect, the medical image analysis device described in the fourth aspect, the computer storage medium described in the fifth aspect, the computer program product described in the sixth aspect, and the chip system described in the seventh aspect are similar to the technical effects achieved by the corresponding technical means in the DICOM file sequence processing method described in the first aspect and the medical image analysis method described in the second aspect, and will not be elaborated further here. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a method for processing a DICOM file sequence according to an embodiment of the present invention; Figure 2A This is a schematic diagram of a process for constructing a sequence DICOM metadata queue according to an embodiment of the present invention; Figure 2B This is a schematic diagram of the structure of a sequence DICOM metadata queue in one embodiment of the present invention; Figure 3 This is a flowchart illustrating a medical image analysis method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the deep learning network processing flow; Figure 5 This is a schematic diagram of the structure of a medical image analysis device provided in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0027] It should be understood that the term "multiple" in this invention refers to two or more. In the description of this invention, unless otherwise stated, " / " indicates "or," for example, A / B can mean A or B; "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist, for example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, to facilitate a clear description of the technical solutions of this application, terms such as "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that terms such as "first" and "second" do not limit the quantity or execution order, and that "first" and "second" do not necessarily imply differences.

[0028] With the continuous development of medical imaging technology, CT, MRI, PET and other computed tomographic medical images have been widely used in clinical diagnosis. These imaging examinations generally use DICOM files for data transmission and storage.

[0029] DICOM is an international standard (ISO 12052) in the field of medical imaging, which defines the format, transmission protocol and storage architecture of medical imaging data, and is widely used in the storage, transmission and sharing of medical imaging information.

[0030] According to the DICOM standard, a single patient imaging examination corresponds to a "Study" object, which typically contains one or more Series objects, each of which usually includes multiple Image objects. Conventional sequences of computed tomography (CT), MRI, PET, and other medical imaging techniques are characterized by continuous scanning of specific anatomical regions of the human body, acquiring a set of images with highly consistent spatial location and scanning direction, conforming to specific patterns. Through advanced 3D reconstruction algorithms, these 2D images can be precisely reconstructed into 3D volumetric images, providing more comprehensive and three-dimensional imaging information for precision medical diagnosis.

[0031] Different sequences correspond to different scanning sites or imaging parameters. In the storage, retrieval, processing, and analysis of medical images, it is crucial to accurately identify the scanning site (such as the head, chest, abdomen, pelvis, etc.) corresponding to each sequence.

[0032] Currently, the identification of scanned areas mainly relies on relevant label information in the DICOM header file, such as study descriptions and series descriptions. However, this descriptive information often lacks standardization due to differences between hospitals, equipment manufacturers, or operators, and may even contain erroneous information, resulting in low accuracy in scanned area identification. Furthermore, relying solely on these text labels for identification cannot fully utilize the rich visual information contained in the images themselves.

[0033] The DICOM file format includes a header information section. This section conforms to the DICOM standard and uses a standard tag list to record and store additional information for each image file, i.e., metadata information, including the examination type, scan parameters, and scanned area. The DICOM header information, or metadata, for different images in a sequence consists of two parts: one part is information common to all images in the sequence, i.e., the parts that are the same for all images in the sequence; the other part is information unique to each image, i.e., each image is different. This differing information is mainly spatial location information, usually arranged and changing according to certain rules, such as an arithmetic progression.

[0034] The DICOM header information set, or metadata set, of different images in a sequence contains a large amount of examination-related information, such as examination type, scan parameters, scan site, and slice spatial location. If this information can be transformed into feature vectors that neural networks can receive and process, especially fixed-length one-dimensional feature vectors, it will make intelligent processing of images within the sequence (such as disease detection and lesion identification) possible.

[0035] Since a sequence may contain hundreds or even thousands of images, directly selecting the DICOM header information set of all images as the feature vector would result in an excessively long feature vector containing a large amount of redundant information. Furthermore, the variable number of images in a sequence also leads to variable feature vector lengths. Deep learning networks, due to their inherent characteristics, require, on the one hand, input feature vectors to be as concise as possible without redundant information, and on the other hand, to have a fixed length. Therefore, current technology limits the processing of DICOM files by deep learning networks.

[0036] Based on the above situation, the present invention provides a method for processing DICOM file sequences, a method for analyzing medical images, a system, a device, a computer storage medium, and a computer program product. It aims to provide a technical solution that can effectively solve the above problems. By removing redundant information from DICOM metadata and establishing a standardized data processing mechanism to unify the feature vector length, it achieves the requirement of simplified and fixed-dimensional input data to adapt to deep learning networks.

[0037] This application embodiment fully utilizes the fact that the DICOM header information of images in a sequence contains sequence-wide information and image characteristic information, and that image characteristic information is usually arranged and changes according to certain rules. It first arranges the images in the sequence into an initial queue, and then selects specific images from designated positions in the initial queue as a "key image file queue," such as the first, middle, and last images in the initial queue. By fixing the length of the "key image file queue," the length of the final feature vector is fixed. Sequence-wide information and image-specific information are extracted from the key image file queue, arranged according to specific rules, finally forming a sequence DICOM metadata queue with a fixed length and fixed information items at specific positions. A deep learning network further transforms the sequence DICOM metadata queue into a more concise feature vector. The core idea is to construct a key image file queue, extract sequence-level DICOM metadata and image-level DICOM metadata from the key image file queue, and construct and output a one-dimensional queue that reflects the overall DICOM metadata numerical characteristics of the sequence in a specified order. It should be noted that DICOM metadata refers to the values ​​corresponding to different tags defined in the DICOM standard. This structured data describes DICOM images and related information, containing a wealth of detailed information about patients, examinations, and images. Through DICOM metadata, doctors and medical staff can gain a more comprehensive understanding of the background and related information of DICOM images, providing important data for medical research and medical information management.

[0038] Figure 1 This is a flowchart illustrating a method for processing DICOM file sequences according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method for processing a DICOM file sequence may include steps S101 to S103.

[0039] S101. Load the DICOM file sequence corresponding to a series of images obtained from the tomographic scan, and arrange them in the scanning order to obtain the initial image file queue.

[0040] First, DICOM files belonging to the same scan series are sorted in scanning order to form an ordered image queue, laying the foundation for subsequent processing. For example, they can be sorted in ascending order according to the values ​​of the DICOM file tags (0020, 0013). Tags (0020, 0013) are instance number tags defined in the DICOM standard, used to identify the image order within the same scan series. A single tomographic scan generates multiple DICOM files, which belong to the same "series" and share the same scanning parameters (such as slice thickness, scan range, etc.). During loading, files belonging to the same series must first be selected (for example, usually identified by the unique identifier (UID) of the series instance, tag (0020, 000e).

[0041] (0020, 0013) correspond to the instance number (InstanceNumber), representing the sequential number of the image in the series (e.g., layer 1, layer 2, etc.). Using InstanceNumber as the key, the DICOM files are sorted in ascending order of value to obtain the initial image file queue. This operation ensures that the order of the image queue is consistent with the physical order of the scan slices (e.g., from head to toe, from left to right), avoiding slice misalignment. For example, a patient's chest CT scan generates 100 DICOM files, with InstanceNumbers from 1 to 100. After sorting, the queue is arranged in the order of 1→2→...→100, corresponding to continuous slices from the apex of the lung to the base of the lung, resulting in the initial image file sequence for this sequence.

[0042] S102. Read DICOM files from a specified position in the initial image file queue to form a fixed-length key image file queue.

[0043] In some possible implementations, the specified position can be the first file, a middle file, or the last file in the initial image file queue. In other possible implementations, the specified position can also be the first file or the last file in the initial image file queue. All are feasible, and this invention does not limit this. As an example, this embodiment of the invention describes an example where the specified position is the first file, a middle file, or the last file in the initial image file queue.

[0044] S103. Read the sequence-level DICOM metadata and image-level DICOM metadata used to represent key features from the key image file queue, arrange them according to preset rules, and obtain a sequence DICOM metadata queue with fixed length and fixed information items at each position.

[0045] In some possible implementations, a sequence of DICOM metadata queues with fixed length and fixed information items at each position is obtained. This includes: filling the sequence of DICOM metadata queues with tag values ​​from the read DICOM file; filling with "" when a tag value is missing, if the data type corresponding to the tag value is text; filling with NaN when the data type corresponding to the tag value is numeric; filling with "" when the data type corresponding to the tag value is enumeration; and normalizing the value when the data type corresponding to the tag value is a value within a continuous range. For example, for the value x of a specific numeric field, the normalized value xnorm is calculated according to the predefined maximum value Xmax and minimum value Xmin using the following formula: The process of building a sequence DICOM metadata queue can be found in [reference]. Figure 2A This includes steps S201 to S207.

[0046] S201. Construct a vector V with an initial length of 0.

[0047] Specifically, initialize the sequence DICOM metadata queue by constructing a sequence DICOM metadata queue with an initial length of 0.

[0048] S202. Write the length N of the initial image file queue to the starting position of vector V. like Figure 2B As shown, the length N of the initial image file queue (i.e., the total number of image files) is written to the beginning position of the queue.

[0049] S203. Read the sequence-level label of the first image file in the key image file queue, perform padding and normalization processing, and add the processed sequence-level DICOM label value to the end of vector V. The number of key image files is M.

[0050] Read the sequence-level DICOM tag value of the first image in the key image file queue, fill and normalize the tag value, and add it to the end of the sequence DICOM metadata queue, such as... Figure 2B As shown.

[0051] When padding and normalizing label values, similar to the process described in step 103, when a label value is missing: if the data type corresponding to the label value is text, fill with ""; if the data type corresponding to the label value is numeric, fill with NaN; if the data type corresponding to the label value is enumeration, fill with ""; if the data type corresponding to the label value is a value within a continuous range, normalize the value. For example, for the value x of a specific numeric field, according to the predefined maximum value Xmax and minimum value Xmin, the formula for calculating the normalized value xnorm is: For example, in some possible implementations, sequence-level DICOM tag values ​​include, but are not limited to, the following tag values.

[0052] (1) Some tag values ​​in the DICOM standard Patient Module.

[0053] Including but not limited to (2) Some label values ​​in the DICOM standard General Study Module Including but not limited to Label Name Tag ID illustrate Study Description (0008,1030) Research Description (3) Some label values ​​in the DICOM standard General Series Module Including but not limited to (4) Some tag values ​​in the DICOM standard Contrast / Bolus Module Including but not limited to Then, add image-level DICOM metadata: starting from the first image file in the key image file queue, read the image-level DICOM tag value of each image file in the key image file queue in sequence, fill and normalize the tag value, and save it to the end of the sequence DICOM metadata queue in sequence, corresponding to steps S204 to S206.

[0054] S204.i=1.

[0055] S205. Read the image-level DICOM tag value of the i-th image file in the key image file queue, perform padding and normalization processing, and add the processed image-level DICOM tag value to the end of vector V.

[0056] When performing padding and normalization, similar to the description in step 103, when a label value is missing: if the data type corresponding to the label value is text, fill with ""; if the data type corresponding to the label value is numeric, fill with NaN; if the data type corresponding to the label value is enumeration, fill with ""; if the data type corresponding to the label value is a value within a continuous range, normalize the value. For example, for the value x of a specific numeric field, according to the predefined maximum value Xmax and minimum value Xmin, the formula for calculating the normalized value xnorm is: S206.i=i+1.

[0057] For example, in some possible implementations, image-level DICOM tag values ​​include, but are not limited to, the tag values ​​described below.

[0058] (1) Some label values ​​in the DICOM standard General Image Module Label Name Tag ID illustrate Patient Orientation (0020,0020) Patient direction Image Position (Patient) (0020,0032) Image location (patient coordinate system) Image Orientation (Patient) (0020,0037) Image orientation (patient coordinate system) Pixel Spacing (0028,0030) Pixel pitch (unit: mm) Samples per Pixel (0028,0002) Number of samples per pixel Photometric Interpretation (0028,0004) Photometric interpretation (e.g., MONOCHROME2) Rows (0028,0010) Image row number Columns (0028,0011) Number of image columns Bits Allocated (0028,0100) Number of bits allocated Bits Stored (0028,0101) Number of bits stored High Bit (0028,0102) Most significant bit Pixel Representation (0028,0103) Pixel representation (e.g., unsigned / signed) (2) Some label values ​​in the DICOM standard CT Image Module For CT images, additional values ​​including but not limited to the following are required. (3) For some label values ​​in the DICOM standard MR Image Module, additional values, including but not limited to the following, are required for MRI images. Label Name Tag ID illustrate Image Type (0008,0008) Image type Acquisition Number (0020,0012) Collection Number Image Position (Patient) (0020,0032) Image position in patient coordinate system Image Orientation (Patient) (0020,0037) Image orientation in patient coordinate system Samples per Pixel (0028,0002) Number of samples per pixel Photometric Interpretation (0028,0004) Photometric interpretation Pixel Spacing (0028,0030) Pixel pitch Slice Thickness (0018,0050) slice thickness Slice Location (0020,1041) Slice position Echo Time (TE) (0018,0081) echo time Repetition Time (TR) (0018,0080) Repeat time Flip Angle (0018,1314) Flip angle Scanning Sequence (0018,0020) Scan sequence Sequence Variant (0018,0021) Sequence variants Echo Train Length (0018,0091) echo train length Inversion Time (0018,0082) Reverse Time (4) For some label values ​​in the DICOM standard PET Image Module, additional values, including but not limited to the following, are required for PET images. S207. Output vector V, which is the sequence DICOM metadata queue.

[0059] In this embodiment, DICOM files are first arranged in scanning order to form an initial image file queue. Then, DICOM files are read from a specified location to form a fixed-length key image file queue. Sequence-level DICOM metafiles and image-level DICOM metafiles are read from the key image file queue to obtain a fixed-length sequence DICOM metadata queue. This effectively eliminates data redundancy in the original sequence, compresses hundreds to thousands of frames of data into a few key frames, significantly reduces the computational burden, and retains core diagnostic information in a targeted manner. The resulting fixed-dimensional sequence DICOM metadata queue provides a data foundation for subsequent input to deep learning networks, which is beneficial for improving the adaptability to deep learning networks and enhancing the accuracy of analysis.

[0060] Figure 3 This is a flowchart illustrating a medical image analysis method according to an embodiment of the present invention, as shown below. Figure 3 As shown, the medical image analysis method may include steps S301 to S306.

[0061] S301. Based on medical images, load the DICOM file sequence corresponding to a series of images obtained from tomographic scans, and arrange them in the scanning order to obtain an initial image file queue.

[0062] S302. Read DICOM files from a specified position in the initial image file queue to form a fixed-length key image file queue.

[0063] S303. Read the sequence-level DICOM metadata and image-level DICOM metadata used to represent key features from the key image file queue, arrange them according to preset rules, and obtain a sequence DICOM metadata queue with fixed length and fixed information items at each position.

[0064] It should be noted that steps S301 to S303 are related to... Figure 1 Steps S101 to S103 are similar; for details, please refer to the previous descriptions of steps S101 to S103, which will not be elaborated here.

[0065] S304. Input the sequence DICOM metadata queue into the deep learning network.

[0066] like Figure 4 As shown, the input layer obtains the sequence DICOM metadata queue. The deep learning network also includes: feature preprocessing and classification layer, feature encoding layer, feature fusion layer, context modeling layer and output layer.

[0067] S305. Trigger the deep learning network to transform the sequence DICOM metadata queue into a fixed-dimensional feature vector.

[0068] This step aims to convert the sequence DICOM metadata queue into a high-dimensional feature vector, effectively capturing the complex patterns and relationships in the DICOM metadata through a deep learning network, thereby generating a DICOM metadata feature representation for subsequent analysis.

[0069] The DICOM metadata encoder employs a deep neural network based on the Transformer architecture, specifically designed to handle the heterogeneous characteristics of medical image DICOM metadata. This encoder can simultaneously process categorical, continuous, textual, and missing value features, generating feature vectors that highly represent sequential DICOM metadata information through multi-level feature extraction and self-attention mechanisms.

[0070] The final output DICOM metadata feature vector is a high-dimensional vector (e.g., 256-dimensional) containing rich semantic information of the sequence DICOM metadata. It can effectively represent key characteristics of medical images such as scanning parameters, equipment information, and patient location, laying the foundation for subsequent medical image analysis.

[0071] This feature vector will serve as an important component of multimodal feature fusion, and will be fused with image information from medical images, enabling its use in scenarios such as automatic identification of scanned areas.

[0072] The following is combined with Figure 4 The functions and processing procedures of each layer in a deep learning network are explained one by one.

[0073] (1) Preprocessing and classification layer The input sequence DICOM metadata queue is first classified and processed according to data type. Categorical features (such as scan parameters, device type, etc.) are transformed through one-hot encoding or embedding technology; continuous features (such as pixel values, scan spacing, etc.) are processed using standardization or Min-Max normalization; textual features (such as inspection description, sequence name, etc.) are processed using sequence encoding methods; missing values ​​are processed using special masking to preserve the semantic meaning of missing information.

[0074] (2) Feature coding layer Categorical features are converted into dense vector representations through an embedding layer (nn.Embedding). Continuous features are processed by linear projection and batch normalization layers to adjust the dimensionality and numerical distribution. Textual features are extracted using a dedicated Transformer encoder to extract semantic information. Masked features are encoded with positional markers to ensure that the model can identify the location and pattern of missing values.

[0075] (3) Feature fusion layer The encoded features of each type are concatenated to form a complete metadata representation. A multi-head attention mechanism is applied to enhance the interaction between features. The attention weights are dynamically adjusted to adjust the importance of each metadata field, highlighting key information related to the identification of the scanned part.

[0076] It should be noted that the preprocessing steps in the preceding steps are used to transform the data into "data in a uniform format." For example, DICOM metadata contains different types of data, among which... Continuous data (e.g., patient age 35 years, scan slice thickness 5 mm) are normalized to the 0-1 range.

[0077] Categorical data (such as gender "male" or equipment type "CT") is encoded using one-hot encoding into numeric vectors such as [1,0] or [0,1].

[0078] Textual data (such as the sequence description "head CT scan") is transformed into a digital vector containing core semantics using the Transformer model.

[0079] Missing values ​​(such as unrecorded contrast agent doses) are indicated to the model with a special label (such as [NaN]) to tell the model "no data here".

[0080] Concatenation can be used to combine processed vectors such as age, gender, device type, and text description in sequence to form a long "feature mosaic". For example: [0.6 (normalized age), 1, 0 (male gender), 0, 1, 0 (CT device), 0.3, 0.7, ... (text semantic vector)] Multi-head self-attention: allowing "multiple experts" to analyze the importance of data. For example, if there are 3 experts, expert 1 focuses on the relationship between "age" and "scanned area": ​​for example, there are more head scans of children, and the weight of the "age" feature will be higher in samples of younger children.

[0081] Expert 2 focuses on the correlation between "equipment type" and "location": CT is often used for the chest, while MRI is often used for the brain, so the characteristics of "equipment CT" have higher weight in chest scans.

[0082] Expert 3 focuses on keywords in the "text description": if the text contains a "header", the weight of "header" related features is increased.

[0083] After each expert conducts an independent analysis, they assign an "importance score" (weight) to each feature. Finally, the opinions of all experts are combined to obtain the final weight distribution.

[0084] The following description uses feature weight adjustment for chest CT scans as an example to illustrate the aforementioned steps. Assume that the DICOM metadata contains the following fields: Age: 50 years old (normalized to 0.5) Gender: Male (One-hot encoding [1,0]) Equipment type: CT (One-hot encoding [0,1,0], assuming there are three types: CT / MRI / PET) Sequence description: "Chest CT Scan" (after processing by Transformer, a semantic vector is obtained, such as [0.8,0.2,0.1,...], where 0.8 represents the semantic meaning of "chest") Layer thickness: 5mm (0.3 after normalization).

[0085] After concatenation, the resulting feature vector is: [0.5,1,0,0,1,0,0.8,0.2,0.1,0.3] (10 dimensions in total). The "expert analysis" process for bullish self-attention includes: Expert 1 (Focusing on age + device): It was found that the probability of chest scans was high among people aged 50 and above using CT equipment, so the weights of "age 0.5" and "CT equipment" were each increased by 0.2.

[0086] Expert 2 (Focusing on text keywords): The keyword "Chest" is directly associated with the chest, so the weight of "0.8 in the text semantic vector" is increased by 0.3.

[0087] Expert 3 (Focusing on layer thickness and location): Chest CT scans commonly use a 5mm slice thickness, so we assign a weight of +0.1 to "slice thickness 0.3".

[0088] Final weight distribution: Age: 0.5 (original value) × 1.2 (weight) = 0.6 CT equipment: 1 (1 in One-hot) × 1.2 = 1.2 The semantic value of the text "chest": 0.8 × 1.3 = 1.04 Other feature weights remain unchanged (such as gender, layer thickness, etc.).

[0089] The significance of multiple heads is understandable: different "heads" can capture correlations across different dimensions (e.g., age-location, text-location, device-location), avoiding information omissions from a single perspective. The model learns through training which feature combinations are most useful for identifying the chest / head / abdomen, and dynamic weights automatically enhance the influence of key features while suppressing irrelevant features (e.g., gender has little impact on chest scans, so its weight can remain unchanged). Compared to traditional methods, which typically rely on manual settings to prioritize text descriptions, the self-attention mechanism allows the model to discover that "sometimes device type is more reliable than text" (e.g., even with an incorrect text description, the CT scanner can still correctly point to the chest).

[0090] (4) Context modeling layer The first feedforward network layer uses linear transformation and GELU activation function to extract high-level feature patterns. Layer normalization maintains training stability and accelerates convergence. The second feedforward network layer further extracts abstract features and applies layer normalization again to ensure stable feature distribution.

[0091] Linear transformation is used to convert data of different units / types into a common unit. For example, if the input data includes "age 50 years old" (numerical), "equipment type CT" (categorical, one-hot encoded as [0,1,0]), and "slice thickness 5mm" (numerical), linear transformation will use mathematical formulas to convert these data into vectors of the same dimension (e.g., all converted to numbers in the range of 0-10). For instance, it converts "50 years old," "CT," and "5mm" into a format suitable for subsequent processing (e.g., age → 0.5, CT → [0,1,0] → 2, slice thickness → 0.5).

[0092] The GELU activation function is used to automatically identify and enhance key data features. Through the formula \(GELU(x)=x\cdot\Phi(x)\) (where \(\Phi(x)\) is the cumulative distribution function of a normal distribution), large numerical features are given higher weights, while small numerical features are weakened or even "filtered out". For example, if the converted value of "equipment type CT" is 2 (relatively large), GELU will make it 2.5 (enhanced); "age 0.5" is relatively small and can remain at 0.5 or be slightly reduced; irrelevant features (such as gender) may be weakened to 0.1.

[0093] The second feedforward network layer is used for abstract feature extraction. It performs "secondary processing" on the data processed in the first stage to extract more complex correlation features. For example, if the first stage obtains "age 0.5", "CT equipment 2.5", and "slice thickness 0.5", the second feedforward network will calculate the combined feature of "CT equipment + slice thickness 5mm" (e.g., 2.5 + 0.5 = 3), find that this is a typical parameter of chest CT, and then generate the abstract feature of "probability of chest CT".

[0094] Layer normalization, used for "standardized calibration" of data, has two main components. The first layer normalization is applied after the first feedforward network to ensure consistent data distribution across different samples within the same batch. For example, if a batch contains 100 samples, each with features such as age, device, and slice thickness, layer normalization calculates the mean and variance of each feature, adjusting the data to a range where the mean is 0 and the variance is 1. The second layer normalization is applied after the second feedforward network to ensure consistent data distribution across different batches, avoiding training fluctuations. For instance, if today's samples are primarily chest CT scans (mean age 45), and tomorrow's samples are primarily head MRI scans (mean age 30), layer normalization will adjust the data for both days to the same distribution range, preventing the model from "learning skewed" due to changes in data distribution.

[0095] The corresponding context modeling layer will be illustrated using the processing of chest CT data as an example. The input data includes: Age: 50 years old (original value) → 0.5 after linear transformation → 0.5 after GELU activation (normal feature, no enhancement); Equipment type: CT (One-hot encoding [0,1,0]) → Linear transformation after 2 → GELU activation after 2.5 (key features, enhancement); Layer thickness: 5mm → 0.5mm after linear transformation → 0.5mm after GELU activation (normal feature); First feedforward network output: [0.5, 2.5, 0.5] → after layer normalization → [0, 1, 0] (assuming the mean of this batch is 1 and the variance is 1); Second feedforward network processing: calculate the combined feature of "equipment CT + slice thickness 5mm": 1 + 0 = 1 → generate abstract feature "chest CT possible" = 0.8; Second layer normalization: If the mean of this feature in the next batch is 0.6 and the variance is 0.2, then adjust it to: ((0.8-0.6)0.2=1) to ensure stable distribution.

[0096] (5) Output layer Global feature pooling summarizes the entire sequence information to generate a fixed-dimensional representation. A fully connected layer adjusts the feature dimension to the required size, batch normalization ensures stable feature distribution, and a Dropout layer (p=0.1) prevents overfitting. Finally, the DICOM metadata feature vector is output with a preset dimension (e.g., 256 or 512).

[0097] The output layer acts as a "packaging stage in the feature processing plant," with pooling used to refine overall features; a fully connected layer standardizes the packaging; batch normalization ensures stable quality; and Dropout enhances anti-interference capabilities. This process accurately identifies medical image scanning areas from different hospitals and using different equipment.

[0098] Global feature pooling is like taking a "family photo" of features. It compresses all features of an entire sequence (such as age, device type, text description, etc.) into a "comprehensive feature package," much like taking a group photo of a group of people, retaining only the overall impression rather than the details of each individual. The input features are 100-dimensional vectors. For example, given 100 features such as "age 50 years old," "device CT," and "slice thickness 5mm," global pooling calculates the average or maximum value of these features, generating a 1-dimensional "comprehensive feature." An analogy: If a class has 50 students, pooling is equivalent to calculating the average height of the entire class, using a single number to represent the overall height feature of the class.

[0099] Fully connected layers are similar to "boxing" features uniformly. Their function includes converting all input features, regardless of their dimensionality, into a fixed dimension (e.g., 256 dimensions), facilitating subsequent processing. For example, if pooling yields 10-dimensional features, a fully connected layer uses mathematical formulas to "stretch" them into 256 dimensions. For instance, 10-dimensional data such as "age + device + slice thickness" can be converted into 256 values, each representing an abstract feature (e.g., "probability of chest scan" or "correlation with CT equipment"). This is analogous to using a cookie cutter; no matter the size of the dough, the resulting cookie will always be a fixed shape.

[0100] Batch normalization is similar to "standardizing" the flavor of features. Its functions include ensuring that the feature distribution of each batch of data is consistent, preventing the model from "learning skewed" due to data fluctuations. For example, if the first batch of data mainly consists of chest CT scans of middle-aged people (mean age 50) and the second batch mainly consists of head MRI scans of young people (mean age 30), batch normalization will adjust the age features to a range with a mean of 0 and a variance of 1. For example, it calculates the mean (e.g., 40 years old) and variance of the "age" feature in a batch of data; it converts each age value into "how far it is from the mean" (e.g., 50 years old → (50-40) / standard deviation, 30 years old → (30-40) / standard deviation); the adjusted age feature distribution is like being "standardized" into the same bowl of soup, with a consistent flavor.

[0101] Dropout layers are like playing "hide-and-seek" with features. Their function includes randomly hiding 10% of the features (p=0.1) when p=0.1, preventing the model from over-relying on certain features. For example, during training, the feature "CT scan equipment" can be intentionally hidden, forcing the model to use other features such as "slice thickness" and "text description" to identify the body part, preventing the model from assuming that "any CT scan indicates a chest area." For instance, one implementation might be: assuming the input has 256 features, during each training iteration, 26 features (256 × 0.1 ≈ 26) are randomly selected and set to 0, while the remaining features are amplified by a factor of 1 / 0.9 ≈ 1.11. For example, the feature value of "CT scan equipment" is 0.9, which might be randomly set to 0, while the feature value of "slice thickness" is 0.7, which is amplified to 0.7 × 1.11 ≈ 0.78.

[0102] The following describes the processing flow of the output layer for a chest CT scan, using the processing steps in the output layer as an example. Input features (before pooling): [0.8 (text "chest"), 0.9 (CT equipment), 0.5 (age 50 years), 0.7 (slice thickness 5mm), ... (total 100 dimensions)]; Global average pooling: Calculate the average value: (0.8 + 0.9 + 0.5 + 0.7 + ...) / 100 = 0.65 (generating a 1D comprehensive feature); Then, fully connected layer (converting to 256 dimensions): Through matrix calculation, the 1D feature is transformed into 256 dimensions, for example: [0.2 (chest probability), 0.7 (CT correlation), 0.1 (year)]. [Age influence),...,0.5 (slice thickness weight)]; Batch normalization: If the mean of the batch data is 0.4 and the variance is 0.2, then "chest probability 0.2" is adjusted to (0.2-0.4) / 0.2=-1, and "CT correlation 0.7" is adjusted to (0.7-0.4) / 0.2=1.5; Dropout (p=0.1): Randomly select 26 features and set them to 0, such as "age influence 0.1" being set to 0, and "slice thickness weight 0.5" being amplified to 0.5×1.11≈0.55; Output result: Finally, a 256-dimensional DICOM metadata feature vector is obtained, which can be used for subsequent fusion with image features to identify the body part.

[0103] S306. Trigger deep learning networks to analyze the metadata semantic information contained in feature vectors and the visual features of medical images.

[0104] In this embodiment, a sequence DICOM metadata queue constructed using a method for processing DICOM file sequences is input into a deep learning network. After processing by the feature preprocessing and classification layers, feature encoding layers, feature fusion layers, context modeling layers, and output layers in the deep learning network, a fixed-dimensional feature vector is obtained. Then, the deep learning network is triggered to analyze the metadata semantic information contained in the feature vector and the visual features of the medical image. This achieves joint analysis of metadata semantic information and the visual features of the medical image, which helps to improve the accuracy of the analysis and makes up for the problem of low accuracy in pure image analysis due to the lack of clinical background.

[0105] This invention also provides a medical image analysis system, comprising: a loading module, a key queue construction module, a metadata queue construction module, and a deep learning network module. The loading module loads a sequence of DICOM files corresponding to a series of images obtained from computed tomography scans and arranges them in the scanning order to obtain an initial image file queue. The key queue construction module reads DICOM files from a specified position in the initial image file queue to form a fixed-length key image file queue. The metadata queue construction module reads sequence-level DICOM metadata and image-level DICOM metadata representing key features from the key image file queue, arranges them according to a preset rule, and obtains a fixed-length sequence DICOM metadata queue with fixed information items at each position. Key features include: scan parameters, device information, and patient location. The deep learning network module converts the sequence DICOM metadata queue constructed by the metadata queue construction module into a fixed-dimensional feature vector and performs analysis based on the metadata semantic information contained in the feature vector and the visual features of the medical images.

[0106] Figure 5 This is a schematic diagram of the structure of a medical image analysis device provided in an embodiment of the present invention. The medical image analysis device 500 includes: a memory 501 and a processor 502. The processor 502 is configured to execute computer-executable instructions. The memory 501 stores one or more computer-executable instructions. When the computer-executable instructions are executed by the processor 502, they implement the steps of any of the processing methods for any pair of DICOM file sequences described in the preceding method embodiments, or implement the steps of any of the medical image analysis methods described in the preceding method embodiments.

[0107] This invention also provides a computer storage medium storing a computer program, which, when executed by a processor, implements any of the embodiments in the preceding method embodiments for processing DICOM file sequences, or any of the embodiments in the preceding method embodiments for medical image analysis.

[0108] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the embodiments in the method embodiments for processing DICOM file sequences as described above, or any of the embodiments in the method embodiments for medical image analysis as described above.

[0109] This invention also provides a chip system applied to an electronic device. The chip system includes one or more processors, which are used to invoke computer instructions to cause the electronic device to execute any method in the method embodiment corresponding to the preceding method for processing DICOM file sequences, or to execute any method in the method embodiment corresponding to the preceding method for medical image analysis.

[0110] It is understood that the beneficial effects achieved by the medical image analysis system, medical image analysis device, computer storage medium, computer program product, and chip system provided above can be referred to the beneficial effects described in the method embodiments, and will not be repeated here.

[0111] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line, DSL) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer, or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0112] The above-described embodiments are optional embodiments provided by this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the technical scope disclosed in this application should be included within the protection scope of this application.

Claims

1. A method for processing DICOM file sequences, characterized in that, include: Load the DICOM file sequence corresponding to a series of images obtained from tomographic scans, and arrange them in the scan order to obtain the initial image file queue; Read DICOM files from a specified position in the initial image file queue to form a fixed-length key image file queue; Sequence-level DICOM metadata and image-level DICOM metadata representing key features are read from the key image file queue, and arranged according to a preset rule to obtain a sequence DICOM metadata queue with fixed length and fixed information items at each position. The key features include: examination type, scanning parameters, scanning site, and patient information.

2. The processing method according to claim 1, characterized in that, The step of reading the DICOM file from a specified position in the initial image file queue includes: Read the first DICOM file, the middle DICOM file, and the last DICOM file sequentially; or, Read the first DICOM file and then the last DICOM file in sequence.

3. The processing method according to claim 1 or 2, characterized in that, The obtained DICOM metadata queue, which has a fixed length and fixed information items at each position, includes: The tag values ​​from the read DICOM file are populated into the sequence DICOM metadata queue; When a label value is missing, fill in "" if the data type of the label value is text; fill in NaN if the data type of the label value is numeric; and fill in "" if the data type of the label value is enumeration. When the data type corresponding to the label value is a numerical value within a continuous range, the value is normalized.

4. A medical image analysis method, characterized in that, include: Based on medical images, a sequence DICOM metadata queue is constructed using the processing method described in any one of claims 1-3; The sequence DICOM metadata queue is input into a deep learning network, which includes a feature preprocessing and classification layer, a feature encoding layer, a feature fusion layer, a context modeling layer, and an output layer. The deep learning network is triggered to transform the sequence DICOM metadata queue into a fixed-dimensional feature vector; The deep learning network is triggered to analyze the metadata semantic information contained in the feature vector and the visual features of the medical image.

5. The medical image analysis method according to claim 4, characterized in that, The feature preprocessing and classification layer classifies the input sequence DICOM metadata queue according to data type; including: Categorical features are transformed using one-hot encoding or embedding techniques, continuous features are processed using standardization or Min-Max normalization, textual features are processed using sequence encoding methods, and missing values ​​are processed using masking.

6. The image analysis method according to claim 4 or 5, characterized in that, The feature fusion layer concatenates the preprocessed and encoded features of each data in the sequence DICOM metadata queue and applies a multi-head self-attention mechanism to dynamically adjust the importance weight of each metadata field.

7. A medical image analysis system, characterized in that, include: The module includes a loading module, a key queue construction module, a metadata queue construction module, and a deep learning network module. The loading module is used to load the DICOM file sequence corresponding to a series of images obtained from tomographic scanning, and arrange them according to the scanning order to obtain an initial image file queue; The key queue construction module is used to read DICOM files from a specified position in the initial image file queue and form a key image file queue of fixed length. The metadata queue construction module is used to read sequence-level DICOM metadata and image-level DICOM metadata representing key features from the key image file queue, arrange them according to preset rules, and obtain a sequence DICOM metadata queue with fixed length and fixed information items at each position; the key features include: scanning parameters, device information and patient location; The deep learning network module is used to convert the sequence DICOM metadata queue constructed by the metadata queue construction module into a fixed-dimensional feature vector; and to analyze the metadata semantic information contained in the feature vector and the visual features of the medical image.

8. A medical image analysis device, characterized in that, The device includes: A processor is configured to execute computer-executable instructions; The memory stores one or more computer-executable instructions, which, when executed by the processor, implement the steps of the method for processing DICOM file sequences as described in any one of claims 1-3, or implement the steps of the medical image analysis method as described in any one of claims 4-6.

9. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for processing DICOM file sequences as described in any one of claims 1-3, or the medical image analysis method as described in any one of claims 4-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for processing DICOM file sequences as described in any one of claims 1-3, or the medical image analysis method as described in any one of claims 4-6.

Images