Multi-mode deep learning based medical image classification device and construction method thereof

Abstract

The invention discloses a deep learning based medical image classification device and a construction method thereof. The device comprises an input module, a coarse classification module, a fine classification module, an integration module and a display module; the coarse classification module comprises a regional convolutional neural network RCNN; the fine classification module comprises a first cyclic convolutional neural network rCNN1 for identifying an original image, a HOG (Histogram of Oriented Gradient) model which converts an image to an HOG, a support vector machine (SVM) for identifying the HOG, a Gaussian mixture model (GMM) and a second cyclic convolutional neural network rCNN2; and the integration module comprises an integrated classifier as a GMM, integrates identification confidence scores of different areas output by the four classifiers into one input vector, inputs the input vector after weighting, and obtains a final identification confidence score of the different areas.

Description

technical field [0001] The invention relates to the fields of deep learning and image recognition, in particular to a medical image classification device based on multi-mode deep learning and a construction method thereof. Background technique [0002] Deep learning has been successfully applied in the field of single image classification and image search, and has developed rapidly in the medical field, such as Google's deep learning of breast cancer CT images, so that the accuracy of breast cancer artificial intelligence screening can reach or exceed tumor doctor. However, due to the large number of unknown parameters of its own model, deep learning requires a huge amount of training data. On the other hand, the cost of labeling medical images is high and the quantity is limited, especially the labeling images of the evolution of the disease process require data of different stages of patients, and often need to collect data in different medical institutions, making labeli...

AI Technical Summary

Problems solved by technology

Although these schemes can carry out deep learning on the characteristics of lung diseases, they do not effectively use the gradient vector information of the diseased tissue itself and the correlation and relative changes with the surrounding healthy tissue (revealing the interconnection of biological phenomena) to improve the performance of deep learning. Specificity and robustness, so especially in the case of limited training image data, even if a trained neural network is obtained, its robustness and accuracy for disease course classification are unsatisfactory, and it is difficult to Approach or surpass doctors in practical application

[0006] Deep learning is based on big data. However, medical images are not easy to obtain massive data due to factors such as information sharing of medical institutions and patient privacy. At the same time, most of the hospital data are terminal patients who have been diagnosed, and patients often change medical institutions as the disease progresses. , so the images and data of the early and complete course of the disease are even rarer, which greatly limits the application of deep learning in the field of medical image recognition

Images