Unlock instant, AI-driven research and patent intelligence for your innovation.

Non-melanoma skin cancer pathological image lesion region detection method based on improved convolutional neural network

A convolutional neural network and pathological image technology, which is applied in the field of lesion area detection in pathological image of non-melanoma skin cancer, can solve the problem of being unable to lock the lesion area and achieve good detection performance

Active Publication Date: 2021-01-26
SHANGHAI UNIV +1
View PDF9 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a non-melanoma skin cancer pathological image lesion detection method based on the improved convolutional neural network, which acts on pathological images of neoplastic skin diseases with complex pathological features, and overcomes the current pathological images of neoplastic skin diseases that only It can classify diseases and cannot lock the lesion area. It can assist doctors to quickly and effectively judge diseases and improve the screening rate of pathological images of neoplastic skin diseases.

Method used

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

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Non-melanoma skin cancer pathological image lesion region detection method based on improved convolutional neural network
  • Non-melanoma skin cancer pathological image lesion region detection method based on improved convolutional neural network
  • Non-melanoma skin cancer pathological image lesion region detection method based on improved convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0038] see figure 1 , a non-melanoma skin cancer pathological image lesion detection method based on an improved convolutional neural network, which includes the following steps:

[0039] 1.1 Obtain pathological images of skin tumors, skin basal cell carcinoma and Bowen's disease, establish an image data set, and label the lesion area;

[0040] 1.2 Construct a deep convolutional neural network model, and establish a convolutional neural network framework suitable for pathological image recognition of neoplastic skin diseases;

[0041] 1.3 Transferring convolutional neural networks pre-trained on large datasets to pathological image recognition for neoplastic skin diseases;

[0042] 1.4 Write scripts for detection of lesion areas in histopathological images, realize pathological image recognition of neoplastic skin diseases, and quickly lock lesion areas.

[0043] This embodiment is based on the improved convolutional neural network lesion region detection method for patholog...

Embodiment 2

[0045] This embodiment is basically the same as Embodiment 1, and the special features are as follows:

[0046] The step 1.1 classifies the pathological images of neoplastic skin diseases obtained clinically by disease classification, and the professional dermatologists mark the lesion area on the pathological images of neoplastic skin diseases, and complete the lesion labeling by covering the lesion area with a rectangle, The apex coordinates of the rectangle and the original image data form a labeled lesion cell tissue image database.

[0047] The step 1.2 improves the deficiencies of the existing YOLOv3 deep convolutional neural network model, and establishes a convolutional neural network framework suitable for pathological image recognition of tumorous skin diseases; after optimizing the 80 layers of the improved YIOLOv3 deep convolutional neural network model A new convolutional layer is added, the convolution kernel is 1×1, and the numbers are 512 and 256 respectively; ...

Embodiment 3

[0052] This embodiment is basically the same as the above-mentioned embodiment, and the special features are as follows:

[0053] figure 1 It is a schematic flow chart of lesion region detection in the present invention, comprising the following steps:

[0054] 1.1 Obtain pathological image samples of skin tumors, skin basal cell carcinoma and Bowen's disease, establish an image data set, mark the lesion area, and form an image database;

[0055] In this example, 800 data sets of basal cell carcinoma and Bowen's disease were selected, of which 600 were used for training and 200 were used for detection;

[0056] For the lesion detection of skin basal cell carcinoma and Bowen's disease, the two indicators of precision and recall are used to measure. The calculation formula of precision is: Among them, TP indicates the number of all real lesions in the detected lesions; FP indicates the number of erroneous lesions in the detected lesions; therefore, the meaning of precision in...

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 Login to View More

Abstract

The invention discloses a non-melanoma skin cancer pathological image lesion region detection method based on an improved convolutional neural network. The method is acted on a tumor skin disease (thebasal cell carcinoma and the Bowen disease) pathological image with complex pathological features. Aiming at the defects of an existing YOLOv3 deep convolutional neural network model, the problems ofdrastic change of number transition and bad training result of convolutional feature maps are improved, and a convolutional neural network framework suitable for pathological image recognition of thetumor skin diseases (the basal cell carcinoma and the Bowen disease) is established. Not only is disease classification of the tumor skin disease pathological image implemented, but also a lesion region is locked, doctors can be assisted in quickly and effectively judging diseases, and the screening rate and accuracy of the tumor skin disease pathological image are improved.

Description

technical field [0001] The invention relates to the field of artificial intelligence target detection in medical images, in particular to a method for detecting lesion regions in non-melanoma skin cancer pathological images based on an improved convolutional neural network. Background technique [0002] There are many types of skin diseases, and there are more than 4,000 known types of diseases. The diagnosis of skin diseases comes from the rich experience of clinicians and perfect laboratory examination evidence. The former depends on the accumulation of knowledge and practice accumulated over the years, and the latter depends on advanced testing methods and methods. Skin pathology is to see the pathological changes of skin epidermis, dermis and other different layers of tissue through a microscope, which extends the field of vision of dermatologists. Clinic and pathology complement each other, and the combination of the two helps doctors to make a correct diagnosis. Howev...

Claims

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

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): A61B5/00
CPCA61B5/444A61B5/7264Y02A90/10
Inventor 张健滔汪鹏宇毕新岭张晓波陈琢周欣薛燕
Owner SHANGHAI UNIV