Urinary calculus detection and classification method based on deep learning and imaging omics

A deep learning and radiomics technology, applied in the field of image processing, can solve the problems of poor algorithm robustness, different sources of CT image equipment, and inability to achieve automatic identification and extraction, so as to overcome errors and improve precision and accuracy. Effect

Pending Publication Date: 2020-06-26
JIANGXI PROVINCIAL PEOPLES HOSPITAL +1
6 Cites 13 Cited by

AI-Extracted Technical Summary

Problems solved by technology

This method also uses manual screening of the region of interest, which cannot be automatically identified and extracted
[0006] It can be seen from the above that how to automatically find the region of interest where the stone is located in the CT image and extract the region of interest is an urgent technical problem at present. At present, the sources of CT ...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Method used

Described step 2) have adopted FasterR-CNN model as the first deep learning model, further as preferred embodiment, Faster R-CNN model is mainly made up of three parts: convolution preprocessing layer; Regional candidate network RPN (RegionProposalNetwork) layer; FastR-CNN classification layer. The combinati...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Abstract

The invention relates to a urinary calculus detection and classification method and system based on deep learning and imaging omics. The method comprises the following steps: firstly, automatically extracting a calculus region of interest from a CT image by utilizing a first deep learning model; and carrying out existence confirmation and primary rough classification on the stone region of interest, and carrying out secondary fine classification on the detected stone region of interest by utilizing a second deep learning model and the extracted iconomics characteristics around the stone in thestone region of interest to obtain a final stone classification result. According to the invention, a plurality of deep learning models and characteristics of calculus imaging omics are combined andutilized, so that an efficient, accurate and full-process automatic calculus classification method is realized, and support is provided for clinical calculus treatment of a subsequent process.

Application Domain

Technology Topic

Image

  • Urinary calculus detection and classification method based on deep learning and imaging omics
  • Urinary calculus detection and classification method based on deep learning and imaging omics
  • Urinary calculus detection and classification method based on deep learning and imaging omics

Examples

  • Experimental program(1)

Example Embodiment

[0093] Example
[0094] The following experiments and evaluation analysis are carried out on the urinary stone detection and classification method based on deep learning and radiomics of the present invention.
[0095] 1. Image acquisition: Obtain routine CT images of urinary calculi with the gold standard.
[0096] 2. On the self-owned labeling system, professional urologists mark the position of the stones on the involved CT images, and label the marked stones according to the gold standard file corresponding to the images. The stones involved in the experiment The types of stones can be divided into two types: calcium stones (mainly calcium oxalate stones) and uric acid stones.
[0097] 3. Divide the data set: the obtained data set is randomly divided into training set, verification set and test set according to the ratio of 6:2:2, wherein the training set and verification set are used to train the stone detection and classification model, and the test set is used for testing performance of the model. At the same time, considering the fact that the training model requires a large amount of data, the training samples are amplified. The preferred methods include commonly used and meaningful data amplification methods such as flipping and random angle rotation.
[0098] 4. Use the training data set to train the urinary stone detection and classification network:
[0099] 1) Stone detection network: According to the number of stone classifications, set the output of the last layer of the network, and set the parameters required in the training process: training batch size, number of epochs, initial learning rate, optimization method adopted, etc.
[0100] 2) Segment the results obtained by the detection network: perform k-means clustering (K=2) on the pixels in the obtained detection frame, and determine the stone area as 1 and the background area as 0 according to the clustering results , and retain the mask of the stone area after segmentation.
[0101] 3) Calculus classification network: Resnet50 is used as the two-stage classification network, the output of the convolutional preprocessing layer before the fully connected layer network of Resnet50 is reshaped into a vector, and then the extracted radiomics features are spliced ​​after it, and the concatenated The obtained feature vector is finally used as the input of the fully connected layer to obtain the classification result.
[0102] 5. Verify the performance of the trained model in the test data set: input the test sample into the model trained in the first stage (that is, the first deep learning model), first obtain the region of interest for the stones detected in the image to be tested, and then Segment the calculus region of interest to obtain its corresponding mask (mask), then extract at least 25 radiomics features corresponding to the calculus region of interest, and finally use the trained two-stage classification network (that is, the second depth learning model) to determine the category of each stone ROI.
[0103] The experimental data is grouped by:
[0104] In 200 test cases, there were 284 stones in total, including 149 calcium stones and 135 uric acid stones.
[0105] Detection phase results:
[0106] Table 1
[0107] Sens FPI Test results 268/284=94.37% 374/200=1.87
[0108] Among them, Sens refers to the detection rate, that is, the number of stones detected by the model/the number of stones included in the actual test set.
[0109] FPI refers to the average number of false positives detected by the model on each case.
[0110] The results of the classification stage (the statistics of the experimental results are carried out on the basis of the detected stones).
[0111] Table 2
[0112]
[0113] The indicators to measure the performance of the classification algorithm are precision rate (precision), recall rate (recall) and F1 value (F1score).
[0114]
[0115] True negative (TN), which is actually the number of samples that negative samples are predicted to be negative samples;
[0116] False positive (FP), which is actually the number of samples that negative samples are predicted to be positive samples;
[0117] False Negative (FN), is actually the number of positive samples predicted as negative samples;
[0118] True positive (TP), which is actually the number of positive samples predicted as positive samples;
[0119] table 3
[0120] stone type precision recall F1 calcium stones 87.68% 88.97% 88.32% Uric acid stones 88.46% 87.12% 87.78%
[0121] It can be seen from the experimental results that the system performs well in distinguishing calcium stones and uric acid stones, and can meet the requirements for distinguishing calcium stones and uric acid stones.
[0122] The content in the above-mentioned method embodiments is applicable to this system embodiment. The specific functions realized by this system embodiment are the same as those of the above-mentioned method embodiments, and the beneficial effects achieved are also the same as those achieved by the above-mentioned method embodiments.
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Similar technology patents

Horizontal honing clamping device

PendingCN114102430AOvercome errorWorkpiece roundness is easyGrinding work supportsPhysicsClip device
Owner:SUZHOU XINNENG PRECISE MACHINERY

Device and method for automatically metering volume of gas on scales of laboratories

PendingCN109489764AOvercome errorOvercoming complexityContainer/cavity capacity measurementEngine displacementEngineering
Owner:CHINA UNIV OF MINING & TECH (BEIJING)

Bone support type double-lantern-ring planting surgery guide plate and using method thereof

PendingCN113974878AOvercome errorGuarantees a precise fitDental implantsDental toolsBiomedical engineeringDental bur
Owner:天津市口腔医院(天津市整形外科医院南开大学口腔医院)

Classification and recommendation of technical efficacy words

  • Improve accuracy and precision
  • Overcome error

High-efficiency precise coal distributing system

InactiveCN103566815AOvercome accuracyOvercome errorMixing methodsSolid fuelsPlant systemProcess engineering
Owner:北京华德创业环保设备有限公司

Temperature missing data filling method and system and electronic equipment

Owner:SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI +1
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products