CN116797869A - Bone tumor image analysis method, system and medium - Google Patents

Bone tumor image analysis method, system and medium Download PDF

Info

Publication number
CN116797869A
CN116797869A CN202310609017.6A CN202310609017A CN116797869A CN 116797869 A CN116797869 A CN 116797869A CN 202310609017 A CN202310609017 A CN 202310609017A CN 116797869 A CN116797869 A CN 116797869A
Authority
CN
China
Prior art keywords
picture sample
data
model
picture
image analysis
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310609017.6A
Other languages
Chinese (zh)
Inventor
杨熙宇
赵毅哲
孙伟
刘满华
杨秀军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai First Peoples Hospital
Original Assignee
Shanghai First Peoples Hospital
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai First Peoples Hospital filed Critical Shanghai First Peoples Hospital
Priority to CN202310609017.6A priority Critical patent/CN116797869A/en
Publication of CN116797869A publication Critical patent/CN116797869A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Abstract

The application provides a bone tumor image analysis method, a bone tumor image analysis system and a bone tumor image analysis medium, which relate to the technical field of image data processing and comprise the following steps: step S1: collecting a picture sample; step S2: preprocessing the picture sample; step S3: constructing a data model and performing cross verification; step S4: training the data model; step S5: inputting the preprocessed picture sample into the data model for calculation, judging the group of the picture sample, and obtaining a group label of the picture sample, namely a picture sample prediction result; step S6: and extracting a characteristic diagram of the last convolution layer of the data model according to the prediction result, calculating to obtain a gradcam diagram, and generating a thermodynamic diagram. The application can effectively assist clinical staff to review the osteosarcoma patient flat tablet, and effectively reduce misjudgment and missed judgment rate of osteosarcoma.

Description

Bone tumor image analysis method, system and medium
Technical Field
The application relates to the technical field of image data processing, in particular to a bone tumor image analysis method, a bone tumor image analysis system and a bone tumor image analysis medium.
Background
Osteosarcoma is a disease with high malignancy, frequently missed diagnosis and misdiagnosis, and the overall incidence of which occupies the first place in teenager bone tumors. The overall prognosis of the osteosarcoma patient is poor, the detection mode comprises means of laboratory examination, imaging examination, pathology examination and the like, and the imaging examination can directly and comprehensively reflect the bone condition, so that the osteosarcoma patient has wide application range. At present, primary screening mainly depends on X-ray films of affected limbs, and the occurrence rate is extremely low, so that judgment is difficult for doctors with insufficient experience. The prior intelligent analysis algorithm of osteosarcoma imaging is mainly applied to judging the prognosis of osteosarcoma, the 5-year survival period of patients, predicting osteosarcoma lung metastasis, and performing preoperative prediction and chemotherapy prediction on early postoperative recurrence.
The prior art does not make a good solution for the early missed judgment and misjudgment of the osteosarcoma, and does not fundamentally improve the judging efficiency of the osteosarcoma. The early detection and early treatment of osteosarcoma are not enough so far, and the clinical first disease local pain is often confused with teenager growth pain, the clinical manifestation heterogeneity is strong, the first diagnosis is mostly carried out in primary hospitals, the first diagnosis examination is also mostly limited to quick and low-cost X-ray flat sheets and the diagnosis level of various doctors in all classes of hospitals is uneven, the diagnosis experience of medical staff is very important, but the characteristics of unobvious characteristics of objective influence factors (shooting precision, image content display integrity, judgment experience, basis and the like) and early internal bone tumor characterization are considered, and the possibility of incapability of detection exists.
Disclosure of Invention
Aiming at the defects in the prior art, the application provides a bone tumor image analysis method, a bone tumor image analysis system and a bone tumor image analysis medium.
According to the bone tumor image analysis method, the bone tumor image analysis system and the bone tumor image analysis medium provided by the application, the scheme is as follows:
in a first aspect, a bone tumor image analysis method is provided, the method comprising:
step S1: collecting a picture sample;
step S2: preprocessing the picture sample;
step S3: constructing a data model and performing cross verification;
step S4: training the data model;
step S5: inputting the preprocessed picture sample into the data model for calculation, judging the group of the picture sample, and obtaining a group label of the picture sample, namely a picture sample prediction result;
step S6: and extracting a characteristic diagram of the last convolution layer of the data model according to the prediction result, calculating to obtain a gradcam diagram, and generating a thermodynamic diagram.
Preferably, the step S1 includes: obtaining more than two image samples of osteoarticular X-ray film data, wherein the image samples comprise osteoarticular X-ray films of patients with osteosarcoma and osteoarticular X-ray films of people with bone joint diseases;
and carrying out data labeling on the data samples, and respectively placing the data with osteosarcoma and the data of healthy people in different folders.
Preferably, the step S2 includes picture sample input and adjustment;
the input of picture samples is that the input picture sample data is read into a matrix with m x n x 3 by using a cv algorithm, the size of each picture sample is m x n pixel points, and each pixel point has three RGB pixel values;
the picture sample adjustment includes: and adjusting the size of the picture sample and normalizing the pixel value of the picture sample.
Preferably, the resizing the picture sample includes: cutting the longer side of the input picture sample from the same side as the shorter side, so that the input size of the picture sample is adjusted to be square;
the normalization processing of the pixel values of the picture samples comprises the following steps: and calculating the average value of all pixel values of the picture sample, subtracting the average value from each pixel value, and dividing the average value by the maximum value of the subtracted result.
Preferably, the step S3 includes: constructing a data model, inputting a picture sample into the data model, setting an output linear matrix as two dimensions, comparing two elements in the output two-dimensional vector, judging that the first numerical generation is relatively large and the second numerical generation is relatively large, and judging that the second numerical generation is relatively large and the second numerical generation is relatively large;
and in the step of cross verification, the picture sample is divided into five parts averagely, one part is selected for testing the model, the other four parts are used for training, one part for testing is converted each time, five models are finally obtained, and the results obtained by the five models are averaged.
Preferably, training the data model in the step S4 includes:
setting model parameters, inputting a plurality of picture samples in batches, wherein the category number is two;
transforming the output two-dimensional matrix into a probability value between 0 and 1 by using a softmax function, wherein the probability value is 1 as a sum of probabilities of two categories;
wherein y is 1 、y 2 The probability of judging healthy and the probability of judging osteosarcoma are respectively represented; x is x 1 、x 2 Respectively representing two primitive elements in the output two-dimensional vector;
using the cross entropy loss function as a loss function, calculating one index from the two output probability values, wherein the smaller the index is, the better the model effect is:
wherein N represents the number of samples of the batch input model; p is p i Representing the probability that the i-th sample is judged to be osteosarcoma; using MURA1.1 numberThe data set is used as pre-training, a model parameter is obtained after the data set is input into a model, and the model parameter is used as an initial parameter of the trained model;
after data of each batch are input, model parameters are adjusted by using an Adam optimizer, so that a loss function is smaller and smaller, and the model is optimized;
five models can be obtained by cross-validation, and the model with the best result in the test sample is taken as the final model.
Preferably, the step S5 includes: and inputting the picture data to be tested into a data model for calculation after data preprocessing, wherein the two element values in the obtained two-dimensional matrix are larger in size, and the category represented by the larger one dimension is the prediction result of the sample.
Preferably, the step S6 includes: in the test process, a pyrach frame is applied to extract the last convolution layer of the data model, and an obtained feature map is obtained;
and (3) inputting the trained data model and the extracted feature images by using a gradcam function, namely outputting to obtain a gradcam image, and covering the gradcam image and the original image to obtain a thermodynamic diagram.
In a second aspect, there is provided a bone tumor image analysis system, the system comprising:
module M1: collecting a picture sample;
module M2: preprocessing the picture sample;
module M3: constructing a data model and performing cross verification;
module M4: training the data model;
module M5: inputting the preprocessed picture sample into the data model for calculation, judging the group of the picture sample, and obtaining a group label of the picture sample, namely a picture sample prediction result;
module M6: and extracting a characteristic diagram of the last convolution layer of the data model according to the prediction result, calculating to obtain a gradcam diagram, and generating a thermodynamic diagram.
Compared with the prior art, the application has the following beneficial effects:
1. the application can effectively assist clinical staff to review the osteosarcoma patient flat tablets, effectively reduce misjudgment and missed judgment rate of osteosarcoma, and bring more floating population to society;
2. the application overcomes subjectivity and personal experience dependence of traditional image diagnosis, realizes homogenization based on X-ray flat plates, and high-level osteosarcoma result analysis, especially early analysis, and further improves prognosis of osteosarcoma patients.
Other advantages of the present application will be set forth in the description of specific technical features and solutions, by which those skilled in the art should understand the advantages that the technical features and solutions bring.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the detailed description of non-limiting embodiments, given with reference to the accompanying drawings in which:
FIG. 1 is a general flow chart of the present application;
FIG. 2 is a diagram of data model construction and cross-validation;
FIG. 3 is a ROC image;
fig. 4 is a visual thermodynamic diagram.
Detailed Description
The present application will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the present application, but are not intended to limit the application in any way. It should be noted that variations and modifications could be made by those skilled in the art without departing from the inventive concept. These are all within the scope of the present application.
The embodiment of the application provides a bone tumor image analysis method, which is shown by referring to fig. 1, and specifically comprises the following steps:
step S1: and collecting a picture sample.
The method specifically comprises the following steps: more than two osteoarticular X-ray film data are acquired, diagnosis is carried out on the osteoarticular X-ray film data by a professional radiologist, tags are marked, the data with osteosarcoma are placed in one folder, and the data of a healthy person are placed in another folder. Osteoarticular X-ray film data are collected by a hospital prescription, and two categories exist, namely, osteoarticular X-ray films of patients with osteosarcoma and osteoarticular X-ray films of healthy people with no diseases of the osteoarticular. The osteoarticular X-ray film data should be stored in the computer in the form of data.
Step S2: and preprocessing the picture sample.
Specifically, the pretreatment includes: inputting and adjusting picture samples; the input of picture samples is that a cv algorithm (computer vision algorithm) should be used to read the input picture sample data into a matrix with m x n x 3, each picture sample has m x n pixel points, and each pixel point has three pixel values of RGB;
the picture sample adjustment includes: and adjusting the size of the picture sample and normalizing the pixel value of the picture sample.
Adjusting the size of the picture sample includes: the longer side of the input picture sample is cropped from the same side as the shorter side so that the input size of the picture sample is adjusted to the square, 224 x 224 size.
The normalization processing of the pixel values of the picture samples comprises the following steps: and calculating the average value of all pixel values of the picture sample, subtracting the average value from each pixel value, and dividing the average value by the maximum value of the subtracted result.
img＝img-img.mean()
img＝img/img.max()
Step S3: and constructing a data model and performing cross-validation.
Referring to fig. 2, input data is input into a resnet34 model, and the resnet34 model is a relatively common convolutional neural network model, which can automatically extract features from an input picture and output the features as a linear matrix. The output linear matrix is set to be 2-dimensional, namely osteosarcoma and normal two categories can be represented, and in the output two-dimensional vector, two elements are compared in size, the first numerical value is larger to judge healthy, and the second numerical value is larger to judge osteosarcoma.
And a cross-validation step, wherein the data are divided into five parts in average, 1 part is selected for a test model, and the other 4 parts are used for training. And each time a transformation is made for one test, 5 models are finally obtained, and the results obtained by the 5 models are averaged.
Step S4: the data model is trained.
Setting model parameters, feeding 16 pictures in batches, wherein the category number is 2, training by using an Adam optimizer, and the initial learning rate is 0.0001.
The 2-dimensional matrix of outputs is transformed to probability values between 0-1 using a softmax function, their sum being 1, to represent the probabilities of the two classes.
Wherein y is 1 、y 2 The probability of judging healthy and the probability of judging osteosarcoma are respectively represented; x is x 1 、x 2 Respectively representing two primitive elements in the output two-dimensional vector;
the cross entropy loss function is used as a loss function, an index is calculated by using the two output probability values, and the smaller the index is, the better the model effect is proved.
Wherein N represents the number of samples of the batch input model; p is p i Representing the probability that the i-th sample is judged to be osteosarcoma; using the MURA1.1 data set downloaded on the network as pre-training, inputting the data set into a model to obtain a model parameter, and using the model parameter as an initial parameter of the trained model; wherein the MURA1.1 dataset was published by team Wu Enda, 14863 studies involving 12173 patients, divided into elbows, and a total of 40561X-rays,Finger, forearm, hand, humerus, shoulder and wrist 7 parts. Marked as normal or abnormal by the radiologist.
After each batch of input, the model parameters are adjusted by using an Adam optimizer, so that the loss function is smaller and smaller, the model is optimized, and the result can be obtained after training for 50 rounds.
Five models can be obtained by cross-validation, and the model with the best result in the test sample is taken as the final model.
Step S5: inputting the preprocessed picture sample into the data model for calculation, judging the group of the picture sample, and obtaining a group label of the picture sample, namely a picture sample prediction result;
specifically, the picture data to be tested is input into a data model for calculation after data preprocessing, and the two element values in the obtained two-dimensional matrix are larger in size, and the category represented by the larger one dimension is the prediction result of the sample.
Step S6: and extracting a characteristic diagram of the last convolution layer of the data model according to the prediction result, calculating to obtain a gradcam diagram, and generating a thermodynamic diagram.
Specifically, in the test process, a pytorch frame is applied to extract the last convolution layer of the data model, and an obtained feature map is obtained;
and inputting the trained data model and the extracted feature map by using a gradcam function, namely outputting to obtain a gradcam map, and covering the gradcam map and the original picture to obtain a thermodynamic diagram for auxiliary diagnosis. The Gradcam function is a function number of the pytorch framework, and can calculate by utilizing the characteristic diagram of the last layer of the model and the type finally judged by the model to obtain the thermodynamic diagram of the focus of the model.
The application also provides a bone tumor image analysis system which can be realized by executing the flow steps of the bone tumor image analysis method, namely, the bone tumor image analysis method can be understood as a preferred implementation mode of the bone tumor image analysis system by a person skilled in the art. The system specifically comprises:
module M1: collecting a picture sample;
module M2: preprocessing the picture sample;
module M3: constructing a data model and performing cross verification;
module M4: training the data model;
module M5: inputting the preprocessed picture sample into the data model for calculation, judging the group of the picture sample, and obtaining a group label of the picture sample, namely a picture sample prediction result;
module M6: and extracting a characteristic diagram of the last convolution layer of the data model according to the prediction result, calculating to obtain a gradcam diagram, and generating a thermodynamic diagram.
Next, the present application will be described in more detail.
Data set:
image data of osteosarcoma patients and healthy people are collected, whether osteosarcoma is detected by radiologists is judged to be the osteosarcoma, the osteosarcoma is classified into two types, the health is marked as 0, the osteosarcoma is marked as 1, and then the data set is randomly classified into 5 parts by adopting a cross-validation mode for cross-validation. Because the health data is less, the data of different positions of the health data in the training set are cut and rotated are enhanced, so that the category distribution is balanced.
And (3) data processing:
since the obtained pictures are different in size, there is also a difference in gray scale. In order to perform batch training in coordination with the network, the picture size needs to be sampled to 224×224 to ensure that the picture sizes are the same. And carrying out normalization processing on the image, subtracting the average value from the pixel value of the image, and dividing the average value by the maximum value to approximate the gray distribution of the image.
Experiment platform:
the experiment is built on a Linux 4.4.0 operating system of an x86_64 processor, an experiment environment adopts a pytorch 1.7.1 framework, python3.8 is used as a programming language, visual studio 1.75.1 is used as an editor, a CPU uses Intel (R) Xeon (R) CPU E5-2620 v4@2.10GHz,GPU to use NVIDIA GeForce RTX 3090, running memory is 24G, and all programs are realized through an open source framework of the pytorch.
Training strategies:
the experiment was trained using a Resnet34 network. Training was performed by using the MURA1.1 dataset, 50 rounds of training were performed on the homemade dataset on the basis of the pre-training parameters, 16 pictures were fed per batch, category number was 2, initial learning rate was 0.0001, and training was performed using the Adam optimizer. During training, the training set data is subjected to horizontal overturn, vertical overturn and data enhancement of random rotation within a (-15 DEG, 15 DEG) range.
The technical result is:
(1) X-ray based osteosarcoma diagnosis
An algorithm for classifying normal X-ray and X-ray of osteosarcoma patients is trained using clinical X-ray images, and ROC images are shown in FIG. 3.
(2) Visual thermodynamic diagram (Grad-Cam diagram)
As shown in fig. 4, the visual diagnostic thermodynamic diagram shows that the final output is not very intuitive as it contains only one positive and negative sample confidence value. And the Grad-CAM method is used for visualizing the main diagnosis basis areas of the positive and negative samples respectively, so that the information richness of auxiliary diagnosis is improved.
The embodiment of the application provides a bone tumor image analysis method, a bone tumor image analysis system and a bone tumor image analysis medium, which are used for diagnosing and identifying X-rays through an algorithm in consideration of the data characteristics of the X-rays of osteosarcoma and visualizing thermodynamic diagrams based on the identification result. Thereby assisting a clinician in providing reference diagnosis opinion and improving the efficiency and accuracy of osteosarcoma X-ray diagnosis.
Those skilled in the art will appreciate that the application provides a system and its individual devices, modules, units, etc. that can be implemented entirely by logic programming of method steps, in addition to being implemented as pure computer readable program code, in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Therefore, the system and various devices, modules and units thereof provided by the application can be regarded as a hardware component, and the devices, modules and units for realizing various functions included in the system can also be regarded as structures in the hardware component; means, modules, and units for implementing the various functions may also be considered as either software modules for implementing the methods or structures within hardware components.
The foregoing describes specific embodiments of the present application. It is to be understood that the application is not limited to the particular embodiments described above, and that various changes or modifications may be made by those skilled in the art within the scope of the appended claims without affecting the spirit of the application. The embodiments of the application and the features of the embodiments may be combined with each other arbitrarily without conflict.

Claims (10)

1. A bone tumor image analysis method, comprising:
step S1: collecting a picture sample;
step S2: preprocessing the picture sample;
step S3: constructing a data model and performing cross verification;
step S4: training the data model;
step S5: inputting the preprocessed picture samples into a data model for calculation, judging the groups of the picture samples, and obtaining group labels of the picture samples, namely, a picture sample prediction result;
step S6: and extracting a characteristic diagram of the last convolution layer of the data model according to the prediction result, calculating to obtain a gradcam diagram, and generating a thermodynamic diagram.
2. The bone tumor image analysis method according to claim 1, wherein the step S1 comprises: obtaining more than two image samples of osteoarticular X-ray film data, wherein the image samples comprise osteoarticular X-ray films of patients with osteosarcoma and osteoarticular X-ray films of people with bone joint diseases;
and carrying out data labeling on the data samples, and respectively placing the data with osteosarcoma and the data of healthy people in different folders.
3. The bone tumor image analysis method according to claim 1, wherein the step S2 includes picture sample input and adjustment;
the input of picture samples is that the input picture sample data is read into a matrix with m x n x 3 by using a cv algorithm, the size of each picture sample is m x n pixel points, and each pixel point has three RGB pixel values;
the picture sample adjustment includes: and adjusting the size of the picture sample and normalizing the pixel value of the picture sample.
4. The bone tumor image analysis method according to claim 3, wherein the resizing the picture sample comprises: cutting the longer side of the input picture sample from the same side as the shorter side, so that the input size of the picture sample is adjusted to be square;
the normalization processing of the pixel values of the picture samples comprises the following steps: and calculating the average value of all pixel values of the picture sample, subtracting the average value from each pixel value, and dividing the average value by the maximum value of the subtracted result.
5. The bone tumor image analysis method according to claim 1, wherein the step S3 comprises: constructing a data model, inputting a picture sample into the data model, setting an output linear matrix as two dimensions, comparing two elements in the output two-dimensional vector, judging that the first numerical generation is relatively large and the second numerical generation is relatively large, and judging that the second numerical generation is relatively large and the second numerical generation is relatively large;
and in the step of cross verification, the picture sample is divided into five parts averagely, one part is selected for testing the model, the other four parts are used for training, one part for testing is converted each time, five models are finally obtained, and the results obtained by the five models are averaged.
6. The bone tumor image analysis method according to claim 1, wherein training the data model in step S4 includes:
setting model parameters, inputting a plurality of picture samples in batches, wherein the category number is two;
transforming the output two-dimensional matrix into a probability value between 0 and 1 by using a softmax function, wherein the probability value is 1 as a sum of probabilities of two categories;
wherein y is 1 、y 2 The probability of judging healthy and the probability of judging osteosarcoma are respectively represented; x is x 1 、x 2 Respectively representing two elements in the output two-dimensional vector;
using the cross entropy loss function as a loss function, calculating one index from the two output probability values, wherein the smaller the index is, the better the model effect is:
wherein N represents the number of samples of the batch input model; p is p i Representing the probability that the i-th sample is judged to be osteosarcoma;
using the MURA1.1 data set as pre-training, and inputting the data set into a model to obtain a model parameter, wherein the model parameter is used as an initial parameter of the trained model;
after data of each batch are input, model parameters are adjusted by using an Adam optimizer, so that a loss function is smaller and smaller, and the model is optimized;
five models can be obtained by cross-validation, and the model with the best result in the test sample is taken as the final model.
7. The bone tumor image analysis method according to claim 1, wherein the step S5 comprises: and inputting the picture data to be tested into a data model for calculation after data preprocessing, wherein the two element values in the obtained two-dimensional matrix are larger in size, and the category represented by the larger one dimension is the prediction result of the sample.
8. The bone tumor image analysis method according to claim 1, wherein the step S6 includes: in the test process, a pyrach frame is applied to extract the last convolution layer of the data model, and an obtained feature map is obtained;
and (3) inputting the trained data model and the extracted feature images by using a gradcam function, namely outputting to obtain a gradcam image, and covering the gradcam image and the original image to obtain a thermodynamic diagram.
9. A bone tumor image analysis system, comprising:
module M1: collecting a picture sample;
module M2: preprocessing the picture sample;
module M3: constructing a data model and performing cross verification;
module M4: training the data model;
module M5: inputting the preprocessed picture sample into the data model for calculation, judging the group of the picture sample, and obtaining a group label of the picture sample, namely a picture sample prediction result;
module M6: and extracting a characteristic diagram of the last convolution layer of the data model according to the prediction result, calculating to obtain a gradcam diagram, and generating a thermodynamic diagram.
10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the bone tumor image analysis method according to any one of claims 1 to 8.
CN202310609017.6A 2023-05-26 2023-05-26 Bone tumor image analysis method, system and medium Pending CN116797869A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310609017.6A CN116797869A (en) 2023-05-26 2023-05-26 Bone tumor image analysis method, system and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310609017.6A CN116797869A (en) 2023-05-26 2023-05-26 Bone tumor image analysis method, system and medium

Publications (1)

Publication Number Publication Date
CN116797869A true CN116797869A (en) 2023-09-22

Family

ID=88043195

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310609017.6A Pending CN116797869A (en) 2023-05-26 2023-05-26 Bone tumor image analysis method, system and medium

Country Status (1)

Country Link
CN (1) CN116797869A (en)

Similar Documents

Publication Publication Date Title
US10417788B2 (en) Anomaly detection in volumetric medical images using sequential convolutional and recurrent neural networks
EP3576100A1 (en) Decision support system for medical therapy planning
US20190138693A1 (en) Methods and apparatus for self-learning clinical decision support
US11651850B2 (en) Computer vision technologies for rapid detection
Thian et al. Deep learning systems for pneumothorax detection on chest radiographs: a multicenter external validation study
US20220180514A1 (en) Methods of assessing lung disease in chest x-rays
Seo et al. A deep learning algorithm for automated measurement of vertebral body compression from X-ray images
Ozdemir et al. Age Estimation from Left-Hand Radiographs with Deep Learning Methods.
Wu et al. A deep learning classification of metacarpophalangeal joints synovial proliferation in rheumatoid arthritis by ultrasound images
Bassi et al. COVID-19 detection using chest X-rays: Is lung segmentation important for generalization?
Li et al. Developing an image-based deep learning framework for automatic scoring of the pentagon drawing test
CN110930373A (en) Pneumonia recognition device based on neural network
Abdikerimova et al. Detection of chest pathologies using autocorrelation functions
Kim et al. Bone age assessment using artificial intelligence in Korean pediatric population: a comparison of deep-learning models trained with healthy chronological and Greulich-Pyle ages as labels
Nigudgi et al. Lung cancer CT image classification using hybrid-SVM transfer learning approach
Guler Ayyildiz et al. Comparison of deep learning methods for the radiographic detection of patients with different periodontitis stages
Xu et al. Identification of benign and malignant lung nodules in CT images based on ensemble learning method
Romanov et al. Artificial intelligence for image-based breast cancer risk prediction using attention
Abdollahi Evaluating LeNet Algorithms in Classification Lung Cancer from Iraq-Oncology Teaching Hospital/National Center for Cancer Diseases
CN116797869A (en) Bone tumor image analysis method, system and medium
Verma et al. Methodologies, Applications, and Challenges of Pneumonia Detection of Chest X-Ray images for COVID-19 using IoT-enabled Deep Learning
Nonthasaen et al. Sex estimation from Thai hand radiographs using convolutional neural networks
Mahapatra et al. Deploying Deep Transfer Learning Models to a Web-App for Sars-Cov-2 Detection Using Chest Radiography Images
López-González et al. Automated chest radiographs triage reading by a deep learning referee network
Fahrudin et al. SkinMate: Mobile-Based Application for Detecting Multi-Class Skin Diseases Classification Using Pre-Trained MobileNetV2 on CNN Architecture

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination