Oral squamous cell carcinoma differentiation degree prediction system

Abstract

The invention discloses an oral squamous cell carcinoma differentiation degree prediction system, and belongs to the field of disease diagnosis and artificial intelligence. The system comprises an input module, a prediction module and an output module. The input module is used for inputting pathological section image data; the prediction module is internally provided with an OfficientNets model; and the output module is used for outputting the slice map which is subjected to region marking according to the differentiation degree. The prediction system is high in prediction accuracy of the differentiation degree of the oral squamous cell carcinoma, and is superior to an existing method for predicting other cancers through deep learning.

Description

Technical field [0001] This application belongs to the diagnosis of disease and artificial intelligence. Background technique [0002] Oral squamous cell carcinoma (OSCC) refers to a malignant tumor that occurs within the oral, squamous cells, cancer cells can occur in gums, hard palate, tongue, cheek mucosa, lips, etc., belong to the viciousness of the head and neck The highest degree, the largest tumor is the most harmful. [0003] OSCC can be divided into high-divided OSCC and low-profile OSCC according to differentiation. High differentiated OSCC's cell proliferation is slow, the transfer is low, the prognosis is good; while low-differentiated OSCC's cell proliferation is fast, easy to transfer, poor treatment, and poor prognosis. [0004] Clinically, the degree of differentiation of OSCC is distinguished primarily on artificially read histoprocess (artificial reading), distinguish between omitacies and high-divided OSCC: The degree of tumor differentiation reflects tumor tis...

AI Technical Summary

Problems solved by technology

Chinese patent application CN111079862A improves the structure of the VGG-f convolutional neural network (that is, removes the last 2 fully connected layers, sets the number of convolution kernels of the fifth convolutional layer to 512, and adds a In the pooling layer, the number of neurons in the fully connected layer is set to 2) to obtain an improved papillary thyroid cancer pathological image classification method. When this method is used for pathological classification of thyroid cancer, the specificity and sensitivity can reach 95% at the same time; The application also compared other existing classification methods, such as Resnet-50, VGG-16, VGG-f, etc., and found that these methods are difficult to balance specificity and sensitivity, and the overall accuracy is not high enough

It can be seen that EfficientNets still cannot take into account the sensitivity and specificity well in the identification of prostate cancer-related pathological slices

[0008] At present, there is no deep learning method for the prediction of the degree of differentiation of OSCC, but according to the recognition of other cancer-related case slices by the deep learning method, the specificity and sensitivity of the deep learning method need to be further improved

Images