Method and device for identifying plasma cell morphological multiple myeloma and AL amyloidosis
By automating the processing of bone marrow smear images and classifying them using deep learning models, the problem of low efficiency in the differential diagnosis of plasma cell diseases has been solved, and a high-precision and rapid identification method has been achieved, which is applicable to the automated identification of multiple myeloma and AL amyloidosis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-22
- Publication Date
- 2026-04-14
AI Technical Summary
Existing technologies suffer from low diagnostic consistency, long processing time, high cost, and low efficiency in the differential diagnosis of plasma cell diseases. Traditional methods are difficult to automate with high throughput and high precision.
By acquiring digital images of bone marrow smears, performing image processing and segmentation, extracting morphological parameters of plasma cells, and using a trained convolutional neural network model for classification, the automated differentiation between multiple myeloma and AL amyloidosis can be achieved.
It enables rapid, objective, and highly accurate automated identification of plasma cell diseases, reduces subjective errors from human observation, improves diagnostic efficiency by nearly 100 times, and meets the needs of high-throughput clinical processing.
Smart Images
Figure CN121860952A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical testing technology, and in particular to a method and apparatus for differentiating plasma cell morphology multiple myeloma from AL amyloidosis. Background Technology
[0002] AL amyloidosis is short for systemic light chain amyloidosis, and can also be simply referred to as AL. Multiple myeloma can be simply referred to as MM. Multiple myeloma and AL amyloidosis are two common plasma cell diseases with similar clinical presentations but drastically different treatment options. MM usually requires chemotherapy or stem cell transplantation, while AL requires targeted removal of light chains. Therefore, accurate differentiation between the two is crucial for clinical decision-making.
[0003] Currently, the diagnosis of hematological diseases mainly relies on pathological examination and molecular biological testing. In the differential diagnosis of plasma cell diseases, traditional pathological microscopy depends on the physician's subjective experience, judging morphological differences in plasma cells in bone marrow smears through microscopic observation, such as nuclear-cytoplasmic ratio and nuclear abnormalities. However, the human eye has limited resolution, making it difficult to quantify subtle morphological characteristics, resulting in low diagnostic consistency. Furthermore, single-sample analysis is time-consuming, typically exceeding 30 minutes, failing to meet the high-throughput requirements of clinical practice. In addition, while molecular detection technologies such as serum free light chain detection (FLC) and immunofixation electrophoresis are considered the gold standard, they have a false negative rate of approximately 15%, and their high cost and long processing time limit their widespread adoption in primary healthcare institutions. These traditional methods still have significant limitations in terms of accuracy, efficiency, and cost.
[0004] Furthermore, while existing technologies attempt to introduce machine learning methods into medical image analysis, the analysis and diagnostic systems still largely rely on manually pre-selected cell regions, failing to achieve end-to-end automated processing and still unable to meet the needs of clinical practice in terms of efficiency and accuracy. Summary of the Invention
[0005] The purpose of this application is to provide a method and device for differentiating multiple myeloma and AL amyloidosis by plasma cell morphology. The method extracts the nuclear-level morphological features of plasma cells through automatic image segmentation and classifies them using a convolutional neural network model optimized by transfer learning. This solves the problems of strong subjectivity and low efficiency of traditional diagnostic methods and realizes rapid, objective and high-precision automated identification of multiple myeloma and AL amyloidosis.
[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for differentiating between multiple myeloma and AL amyloidosis based on plasma cell morphology, comprising: acquiring an original digital image of a bone marrow smear; performing image processing on the original digital image to generate plasma cell segmentation data, wherein the plasma cell segmentation data is used to identify the plasma cell regions and nucleus regions of multiple plasma cells; calculating morphological parameters of the plasma cells in the bone marrow smear based on the plasma cell segmentation data to generate a feature dataset, wherein the morphological parameters include nuclear-cytoplasmic ratio and nuclear roundness; inputting the feature dataset into a trained discrimination classification model; receiving and outputting the discrimination result output by the discrimination classification model, wherein the discrimination result includes multiple myeloma, AL amyloidosis, and no disease.
[0007] Optionally, the discrimination and classification model is a deep learning model based on a convolutional neural network.
[0008] Optionally, the convolutional neural network is a ResNet18 architecture; and the discriminative classification model is trained by adjusting the parameters of the top classification layer through transfer learning.
[0009] Optionally, the step of performing image processing on the original digital image to generate plasma cell segmentation data includes: preprocessing the original digital image to obtain preprocessed image data, wherein the preprocessing includes at least denoising and color normalization; and performing image segmentation on the preprocessed image data using an edge detection algorithm to generate the plasma cell segmentation data.
[0010] Optionally, the training steps of the trained discrimination and classification model include: obtaining a training dataset, which includes multiple sets of stained bone marrow smear digital images, labeled plasma cell locations, and disease labels, wherein the plasma cell locations and disease labels are labeled by a pathologist; and iteratively training the discrimination and classification model based on the training dataset using a stochastic gradient descent optimizer and a cross-entropy loss function.
[0011] Optionally, the method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis further includes: generating a diagnostic report based on the differentiation results; wherein the diagnostic report includes the differentiation results and a distribution heatmap based on the generation of multiple plasma cells.
[0012] Secondly, this application provides a device for differentiating between plasma cell morphology multiple myeloma and AL amyloidosis, comprising: an image acquisition module configured to acquire a raw digital image of a bone marrow smear; an image processing module configured to perform image processing on the raw digital image to generate plasma cell segmentation data, wherein the plasma cell segmentation data is used to identify plasma cell regions and nucleus regions of multiple plasma cells; a feature extraction module configured to calculate morphological parameters of the plasma cells in the bone marrow smear based on the plasma cell segmentation data to generate a feature dataset, wherein the morphological parameters include nuclear-cytoplasmic ratio and nuclear roundness; a model identification module configured to input the feature dataset into a trained identification classification model; and a result output module configured to receive and output the identification results output by the identification classification model, wherein the identification results include multiple myeloma, AL amyloidosis, and no disease.
[0013] Optionally, the device for differentiating plasma cell morphology multiple myeloma from AL amyloidosis further includes: a report generation module configured to generate a diagnostic report based on the identification results; wherein the diagnostic report includes the identification results and a distribution heatmap based on the generation of multiple plasma cells.
[0014] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis as described above.
[0015] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis as described above.
[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method and apparatus for differentiating plasma cell morphology from multiple myeloma and AL amyloidosis. It automatically generates segmentation data of plasma cells and their nuclei through image processing, and calculates key morphological parameters such as nucleoplasmic ratio and nuclear roundness, reducing subjective errors and measurement fluctuations in human observation and laying a reliable data foundation for subsequent accurate classification. By inputting standardized feature datasets into a trained differential classification model for inference, it can deeply explore the complex nonlinear relationship between morphological features and disease types, ultimately outputting high-precision identification results. By automatically generating a diagnostic report containing the identification results and a plasma cell distribution heatmap, it not only provides qualitative conclusions but also presents quantitative evidence in a visual form, facilitating rapid understanding and adoption by physicians and enabling seamless integration with hospital information systems. Therefore, by automatically segmenting, extracting features, and classifying raw digital images of bone marrow smears, the analysis time for a single sample is significantly shortened, improving efficiency by nearly 100 times, achieving high-throughput batch processing in clinical practice. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is an application environment diagram of a method for differentiating multiple myeloma and AL amyloidosis based on plasma cell morphology in one embodiment of this application. Figure 2 A flowchart illustrating a method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis, provided in an embodiment of this application; Figure 3 This is a detailed flowchart illustrating the training steps of a classification model in one embodiment of this application. Figure 4 for Figure 2 A schematic diagram of the image processing refinement process in step S12; Figure 5 A schematic diagram of the functional modules of a device for differentiating plasma cell morphology multiple myeloma from AL amyloidosis, provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation
[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0021] The method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be set up independently, integrated into server 102, or placed in the cloud or on another server. Terminal 101 can send the raw digital image of a bone marrow smear to server 102. Server 102 receives the raw digital image of the bone marrow smear, performs image processing on the raw digital image, and generates plasma cell segmentation data. The plasma cell segmentation data is used to identify multiple plasma cell regions and their corresponding nucleus regions. Based on the plasma cell segmentation data, morphological parameters of each of the multiple plasma cells are calculated to generate a feature dataset. The morphological parameters include at least the nuclear-cytoplasmic ratio and nuclear roundness. The feature dataset is input into a trained discrimination and classification model, and the model receives and outputs the discrimination result indicating multiple myeloma or AL amyloidosis. Server 102 can feed back the obtained discrimination result to terminal 101. In addition, in some embodiments, the method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis can also be implemented by the server 102 or the terminal 101 alone. For example, the terminal 101 can directly process the original digital image of the bone marrow smear to be processed, or the server 102 can obtain the original digital image of the bone marrow smear from the data storage system.
[0022] The terminal 101 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 102 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.
[0023] In one exemplary embodiment, such as Figure 2As shown, a method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 We will use server 102 as an example to illustrate this.
[0024] like Figure 2 As shown, a method for differentiating plasma cell morphology from multiple myeloma and AL amyloidosis includes the following steps: S11, Obtain the raw digital image of the bone marrow smear; S12, Image processing is performed on the original digital image to generate plasma cell segmentation data. The plasma cell segmentation data is used to identify the plasma cell regions and nucleus regions of multiple plasma cells. S13. Based on plasma cell segmentation data, calculate the morphological parameters of plasma cells in bone marrow smears to generate a feature dataset, wherein the morphological parameters include nuclear-cytoplasmic ratio and nuclear roundness. S14, Input the feature dataset into the trained discriminative classification model; S15, Receive and output the identification results output by the identification classification model, including multiple myeloma, AL amyloidosis and no disease; S16. Based on the identification results, generate a diagnostic report; the diagnostic report includes the identification results and a distribution heatmap based on multiple plasma cells.
[0025] This application presents a method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology. It automatically generates segmentation data of plasma cells and their nuclei through image processing, and calculates key morphological parameters such as the nucleus-to-cytoplasm ratio and nuclear roundness. This reduces subjective errors and measurement fluctuations from human observation, laying a reliable data foundation for subsequent accurate classification. By inputting a standardized feature dataset into a trained deep learning model for inference, it can deeply explore the complex nonlinear relationship between morphological features and disease types, ultimately outputting high-precision identification results. The automatically generated diagnostic report, including the identification results and a plasma cell distribution heatmap, not only provides qualitative conclusions but also presents quantitative evidence in a visual form, facilitating rapid understanding and adoption by physicians and seamless integration with hospital information systems. Therefore, by automatically segmenting, extracting features, and identifying the raw digital images of bone marrow smears, end-to-end automated processing is achieved, significantly shortening single-sample analysis time and improving efficiency by nearly 100 times, enabling high-throughput batch processing in clinical settings.
[0026] In practice, the raw digital image of a bone marrow smear refers to the image obtained by scanning or photographing a stained bone marrow smear sample using a digital imaging device, including but not limited to a digital microscope system, a fully automated bone marrow smear scanner, and a conventional optical microscope equipped with a high-resolution camera. The raw digital image ensures the authenticity and clinical applicability of the image source. It may contain inherent noise, uneven illumination, or staining differences during the imaging process. Therefore, the raw digital image can be processed using step S12 of the plasma cell morphology differentiation method for multiple myeloma and AL amyloidosis described in this application.
[0027] In practice, the feature dataset is a structured data table or vector set, where each row represents all the morphological parameters of a plasma cell, such as nucleoplasm, nuclear roundness, cell area, etc., thereby achieving a quantitative description of the morphology of the cell population.
[0028] In practice, morphological parameters also include cell area, cell perimeter, nuclear area, and nuclear perimeter.
[0029] In practice, the discrimination and classification model is a deep learning model based on a convolutional neural network (CNN), specifically the ResNet18 architecture. The model is trained by adjusting the parameters of the top classification layer through transfer learning. This means that the general image feature extraction capabilities obtained by the ResNet18 architecture model serve as prior knowledge. This knowledge is then transferred to the specific domain of bone marrow smear images. By retaining the underlying general feature extraction layer and adjusting the parameters of the top classification layer, the model can quickly and efficiently focus on learning the subtle differences in plasma cell morphology most relevant to the differentiation between MM and AL amyloidosis. This effectively overcomes the bottleneck of training on small samples in medical imaging, transforming generalization ability on large datasets into high-precision discrimination ability for specific tasks.
[0030] like Figure 3 As shown, the training steps for the discrimination and classification model include: S21. Obtain the training dataset, which includes multiple sets of stained bone marrow smear digital images, labeled plasma cell locations, and disease labels. The plasma cell locations and disease labels are labeled by pathologists. S22, based on the training dataset, uses a stochastic gradient descent optimizer and a cross-entropy loss function to iteratively train the discrimination and classification model.
[0031] Understandably, the training dataset includes multiple sets of stained bone marrow smear digital images and labeled plasma cell locations and disease labels. The high-quality, supervised training samples, constructed through pathologist annotations, help ensure the accuracy of the classification model's learning objectives. Furthermore, the combination of a stochastic gradient descent optimizer and a cross-entropy loss function ensures stable convergence during model training, effectively finding optimal solutions in the complex parameter space. This ensures the classification model fully learns the complex mapping relationship from morphological features to disease labels, driving model parameters to update in a direction that improves the classification accuracy of MM and AL amyloidosis. In practical applications, the aforementioned classification model achieves a high accuracy rate of 90.59%, and its AUC (area under the curve) is as high as 98.18%, indicating a low error rate and high reliability.
[0032] In specific implementation, such as Figure 4 As shown, step S12, which involves image processing of the original digital image to generate plasma cell segmentation data, includes: S121, preprocess the original digital image to obtain preprocessed image data, wherein the preprocessing includes at least denoising and color normalization; S122, use an edge detection algorithm to segment the preprocessed image data to generate plasma cell segmentation data.
[0033] Understandably, denoising aims to eliminate random noise introduced during microscopic imaging, providing a clear image foundation for subsequent edge detection. Specifically, denoising can employ Gaussian filtering or median filtering algorithms to smooth the image and suppress noise. Color normalization is used to eliminate color differences caused by different staining times and reagent batches, ensuring that image data from different sources are processed in a unified color space. This makes the extracted morphological parameters have reliable comparative significance, eliminates interference from non-biological factors, and solves the technical problems of inconsistent quality and poor feature extraction stability in original bone marrow smear images caused by noise and staining differences. It achieves image data standardization, which is an important step in achieving standardized diagnosis. Based on this, edge detection algorithms are specifically used to capture the contour features of plasma cells and their nuclei, thereby accurately separating them from the complex image background.
[0034] In practice, diagnostic reports can include: statistical summaries of various morphological parameters of multiple plasma cells, such as mean, standard deviation, maximum, and minimum values; classification confidence or probability values output by the differential classification model; comparative analysis with a typical case database; and a text-based diagnostic conclusion conforming to clinical standards, such as a high probability of multiple myeloma (MM). Diagnostic reports can also be displayed through a graphical user interface (GUI) or directly transmitted to a hospital information system (HIS) or laboratory information system (LIS) for seamless integration with clinical workflows.
[0035] For example, the classification model in this application is a deep learning model based on a convolutional neural network, which is based on the ResNet18 architecture. When the ResNet18 architecture is replaced by the ViT model, the ViT model is a deep learning model that directly applies the Transformer architecture, originally used in the field of natural language processing, to image sequences. Its core idea is to split the image into a series of image patches and understand global information through a self-attention mechanism. Since its attention mechanism relies heavily on massive amounts of data for training, it shows a significant tendency to overfit on limited medical image datasets. The model pays too much attention to irrelevant noise features, resulting in a significant drop in its accuracy on independent test sets. In addition, although the global attention map of ViT can highlight the general area, it is difficult to accurately locate local subtle morphological features that are crucial for identification, such as nuclear membrane irregularities. This reduces the reliability and clinical interpretability of the model's decision-making process. When the ResNet18 architecture is replaced by the MobileNet model, the depthwise separable convolutional structure used in the latter, in pursuit of computational speed, can quickly capture the general outline, but may lose or fail to fully resolve the kernel-level subtle features crucial for identification. Essentially, this sacrifices the depth and richness of feature extraction. The extracted features are insufficient to fully characterize the complex morphological differences between MM and AL plasma cells, resulting in an upper limit of the model's overall discrimination ability (AUC) of only 94.2%, significantly lower than the 98.18% achieved by the ResNet18 architecture, failing to meet the stringent high-precision requirements of clinical auxiliary diagnosis. This application combines the ResNet18 architecture with a transfer learning strategy, a unique technical solution proposed for the task of identifying plasma cell morphology in medical images, which simultaneously achieves high accuracy, strong generalization ability, and good interpretability.
[0036] For example, such as Figure 5 As shown, a device for differentiating plasma cell morphology multiple myeloma from AL amyloidosis is provided, comprising: an image acquisition module 301, an image processing module 302, a feature extraction module 303, a model identification module 304, a result output module 305, and a report generation module 306.
[0037] Image acquisition module 301 is configured to acquire raw digital images of bone marrow smears; Image processing module 302 is configured to perform image processing on the original digital image to generate plasma cell segmentation data, which is used to identify the plasma cell regions and nucleus regions of multiple plasma cells. The feature extraction module 303 is configured to perform morphological parameter calculation of plasma cells in bone marrow smears based on plasma cell segmentation data to generate a feature dataset, wherein the morphological parameters include nuclear-cytoplasmic ratio and nuclear roundness; The model discrimination module 304 is configured to perform the input of the feature dataset into the trained discrimination classification model; The output module 305 is configured to receive and output the identification results from the identification classification model, including multiple myeloma, AL amyloidosis, and no disease.
[0038] The report generation module 306 is configured to generate a diagnostic report based on the identification results; wherein the diagnostic report includes the identification results and a distribution heatmap generated based on multiple plasma cells.
[0039] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 6 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores raw digital images of bone marrow smears. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis.
[0040] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0041] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0042] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0043] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0044] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0045] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0046] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0047] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0048] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology, characterized in that, include: Obtain the raw digital image of the bone marrow smear; The original digital image is processed to generate plasma cell segmentation data, which is used to identify the plasma cell regions of multiple plasma cells and the nucleus regions of the plasma cells. Based on the plasma cell segmentation data, the morphological parameters of the plasma cells in the bone marrow smear are calculated to generate a feature dataset, wherein the morphological parameters include the nuclear-cytoplasmic ratio and nuclear roundness; The feature dataset is then input into the trained discriminative classification model; Receive and output the identification results output by the identification classification model, the identification results including multiple myeloma, AL amyloidosis and no disease.
2. The method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology according to claim 1, characterized in that, The identification and classification model is a deep learning model based on convolutional neural networks.
3. The method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology according to claim 2, characterized in that, The convolutional neural network is based on the ResNet18 architecture; and the discrimination classification model is trained by adjusting the parameters of the top classification layer through transfer learning.
4. The method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology according to claim 1, characterized in that, The step of performing image processing on the original digital image to generate plasma cell segmentation data includes: The original digital image is preprocessed to obtain preprocessed image data, wherein the preprocessing includes at least noise reduction and color normalization; The preprocessed image data is segmented using an edge detection algorithm to generate plasma cell segmentation data.
5. The method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology according to claim 1, characterized in that, The training steps of the trained discrimination and classification model include: Obtain a training dataset, which includes multiple sets of stained bone marrow smear digital images, labeled plasma cell locations, and disease labels, wherein the plasma cell locations and disease labels are labeled by a pathologist. Based on the training dataset, the discriminative classification model is iteratively trained using a stochastic gradient descent optimizer and a cross-entropy loss function.
6. The method for differentiating multiple myeloma from AL amyloidosis based on plasma cell morphology according to claim 1, characterized in that, Also includes: Based on the identification results, a diagnostic report is generated; wherein the diagnostic report includes the identification results and a distribution heatmap based on the multiple plasma cells.
7. A device for differentiating between plasma cell morphology-related multiple myeloma and AL amyloidosis, characterized in that, include: The image acquisition module is configured to acquire raw digital images of bone marrow smears; The image processing module is configured to perform image processing on the original digital image to generate plasma cell segmentation data, wherein the plasma cell segmentation data is used to identify the plasma cell regions of multiple plasma cells and the nucleus regions of the plasma cells; The feature extraction module is configured to perform calculation of morphological parameters of the plasma cells in the bone marrow smear based on the plasma cell segmentation data to generate a feature dataset, wherein the morphological parameters include nuclear-cytoplasmic ratio and nuclear roundness; The model discrimination module is configured to input the feature dataset into a trained discrimination classification model; The results output module is configured to receive and output the identification results from the identification classification model, including multiple myeloma, AL amyloidosis, and no disease.
8. The device for differentiating multiple myeloma and AL amyloidosis based on plasma cell morphology according to claim 7, characterized in that, Also includes: The report generation module is configured to generate a diagnostic report based on the identification results; wherein the diagnostic report includes the identification results and a distribution heatmap based on a plurality of plasma cells.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis as described in any one of claims 1-6.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for differentiating plasma cell morphology multiple myeloma from AL amyloidosis as described in any one of claims 1-6.