Weak supervision automatic diagnosis and grading system based on similarity feature map

Abstract

The invention discloses a weak supervision automatic diagnosis and classification system based on a similarity feature map, and relates to the technical field of deep learning, computer vision and medical imaging. According to the method, the similarity feature map is applied to diagnosis and grading of the brain glioma, a lesion area in brain magnetic resonance imaging (MRI) does not need to be marked, and whether a patient suffers from the glioma or not is judged through the MRI under the condition of less manual marking. If the patient suffers from the glioma, lesion areas are automatically extracted and classified, so that the classification of the glioma is judged. According to the glioma diagnosis and grading system, the analysis and diagnosis precision can reach or even exceed the level of international experts, on one hand, doctors can be assisted to make more accurate medical decisions, the misdiagnosis or excessive diagnosis rate is reduced, the treatment effect and the survival rate of patients are improved, and on the other hand, the glioma diagnosis and grading system can save a large amount of manpower and reduce the cost. And meanwhile, a series of diagnosis problems caused by non-uniform distribution of medical resources are relieved.

Description

technical field [0001] The invention relates to the technical fields of deep learning, computer vision and medical imaging, and in particular relates to a weakly supervised automatic diagnosis and classification system based on a similarity feature map. Background technique [0002] Medical diseases mainly rely on MRI to judge. The National Brain Tumor Society report mentioned that there were about 79,270 brain tumor patients in 2017, and the survival rate was 34.70%. Among them, 16,947 people were diagnosed with malignant brain tumors and died of the disease. . In 2016 it was also mentioned that 77,670 people had brain tumors, with a survival rate of 34.40%. Although the survival rate increased by 0.30% due to improved medical standards, the number of people with brain tumors still increased by 2%, which shows the increasing importance of diagnosing tumors at an early stage. Glioma is the most common primary intracranial tumor, accounting for 40% to 50% of intracranial tu...

AI Technical Summary

Problems solved by technology

[0004] In addition, it is worth noting that glioma is not only difficult to treat, but also often diagnosed as other neurological diseases by neurosurgeons before diagnosis, which delays the best treatment time for patients to a certain extent

This is mainly because in the clinical environment, brain magnetic resonance imaging (MRI) is usually used to diagnose whether a patient has glioma. This diagnostic mode based on image analysis is completely dependent on the allocation and supply relationship of medical resources. In some medical In areas with weak resources, the number of doctors is seriously insufficient, and imaging diagnosis depends on human knowledge and experience, resulting in variable and inaccurate analysis results

Images