Lung disease auxiliary diagnosis cloud platform based on deep learning
A technology for assisting diagnosis and pulmonary diseases, applied in neural learning methods, medical automated diagnosis, computer-aided medical procedures, etc., can solve the problems of long training cycle for radiologists, lack of products that can be used by the public, and poor scalability. To achieve the effect of strong scalability and portability
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0050] The schematic diagram of the embodiment of the lung disease aided diagnosis cloud platform in the present invention is as follows: figure 1 shown. The cloud platform can be mainly divided into two parts: server and client. The client includes the presentation layer, including the web (web page) and WeChat applets. The interfaces of the two types of clients are different, but the user data and the permissions and functions owned by the user are consistent. The two are not mutually exclusive and can be operated at the same time. . From top to bottom, the server is the data access layer, business logic layer and control layer. The control layer includes the verification and processing of the Ajax asynchronous request submitted by the web page and the request of the WeChat applet client via the WeChat server. If it passes, the control layer will feed back to the business logic layer. If the verification error or request is abnormal, Then the control layer makes a corres...
Embodiment 2
[0063]As described in Example 1, the 12 kinds of lung disease diagnosis weight data files used by the analysis module in the platform server business logic layer are 110,000 chest radiograph data sets released by the National Institutes of Health (NIH) through the Xception neural network. The result after network training, the weight data file is obtained through the following steps.
[0064] The number distribution of each pathology in the original data set is relatively unbalanced, and the normal cases without disease account for the majority (59.483%). Therefore, in this example, the example of No finding (normal) is discarded at first to improve the ability of disease diagnosis, and since there are only 227 examples of Hernia (hernia), the present invention only selects a total of 12 pathologies with a quantity of more than 1000.
[0065] Furthermore, since there are 12 kinds of lung diseases, the present invention marks the data set images with numbers from 1 to 12 as lab...
Embodiment 3
[0080] As described in Embodiment 1, the pulmonary nodule position diagnosis weight data file used by the analysis module in the business logic layer of the platform server is the result after training using the frontal chest X-ray image provided by the Japanese Society of Radiology Technology (JSRT). The weight data file is obtained through the following steps.
[0081] Part of the data in this example selects a total of 147 effective data in the data set (referring to chest films that can be opened and used normally and are not damaged), each of which has one and only one pulmonary nodule, and corresponds to the location of the pulmonary nodule The pixel coordinates of the center (unit: pixel) and the radius of the lung nodule (unit: mm).
[0082] Further, this example extracts the selected 147 chest radiographs in IMG format and saves them as black and white images in three-channel jpg format. This example converts the lung nodule radius data stored in the txt file, conver...
PUM

Abstract
Description
Claims
Application Information

- R&D
- Intellectual Property
- Life Sciences
- Materials
- Tech Scout
- Unparalleled Data Quality
- Higher Quality Content
- 60% Fewer Hallucinations
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com