Leucorrhea sample automatic detection method and system based on deep learning and storage medium
A technology of automatic detection and deep learning, applied in the field of medical microscopic image processing, can solve the problems of slow detection speed and low accuracy, and achieve the effect of reducing professionalism, ensuring accuracy and improving detection speed.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0038] Such as Figure 7 As shown, this implementation discloses a method for automatic detection of leucorrhea samples based on deep learning, including the following steps:
[0039] Collect leucorrhea sample pictures, and mark the categories and positions of active ingredients in the leucorrhea sample pictures;
[0040] Build a leucorrhea detection model based on the neural network framework of Faster R-CNN, and use ResNet-50 as the backbone feature network of the leucorrhea detection model;
[0041] The marked leucorrhea sample picture is used to train the leucorrhea detection model, and the leucorrhea detection model is used to detect the leucorrhea sample.
[0042] In addition, in this embodiment, a computer system is also disclosed, including a memory, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the computer program, the steps of any of the above methods are implemented. .
[0043] In addition, in...
Embodiment 2
[0046]Embodiment 2 is a preferred embodiment of Embodiment 1, and its difference from Embodiment 1 is that the specific steps of the automatic detection method for leucorrhea samples based on deep learning are refined:
[0047] The present invention is written based on the Python language, and mainly uses Tensorflow, the mainstream deep learning library. The version is 1.13.1, and the Python version is 3.6 or above. If the computer is equipped with a GPU, the training speed will be improved. It is recommended to use light The magnitude of the software VS Code.
[0048] The target detection algorithm model adopted is the Faster R-CNN algorithm in target detection, and the backbone feature extraction network is selected as ResNet-50. For the specific network structure and calculation process, see figure 2 , the algorithm internally adjusts the input image to a uniform size, so there is no specific limit on the size of the input image.
[0049] Specific implementation steps:
...
Embodiment 3
[0068] Embodiment three is the preferred embodiment of embodiment two, and its specific content is as follows:
[0069] In this example, in the experiment, 484 single-cell data sets were used to build and train the classification network, and the training set, verification set and test set were divided into 8:1:1, and the average accuracy rate was 99.31%, which is considered acceptable. meet clinical needs.
[0070] In the experiment, 744 sample images were trained and 124 sample images were tested. In the test results, the mAP of epithelial cells and white blood cells reached 97.13%, which is considered to meet the clinical needs.
[0071] Among them, the automatic detection process of leucorrhea sample in this embodiment is as follows:
[0072] Step 1: Acquire Dataset
[0073] Step 2: Label the data set with labelimg, you can label the cells that need to be detected, see the labeling process figure 1 ; The present invention only labels leukocytes and epithelial cells. Epi...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More