DICOM medical image sequence classification method for artificial intelligence auxiliary diagnosis

Abstract

The invention relates to a DICOM medical image sequence classification method for artificial intelligence auxiliary diagnosis. According to the method, DICOM data passes through five layers of classifiers in sequence; finally, a classification result of a specific sequence is obtained, through a mixed strategy of multiple sequence classifiers, a user operates for a period of time, each result is updating and supplementing of the classifiers, learning training of the auxiliary diagnosis system is achieved, and recognition errors of a single method are corrected.

Description

technical field [0001] The invention relates to a DICOM medical image sequence classification method for artificial intelligence-assisted diagnosis. Background technique [0002] The auxiliary diagnosis of medical images based on artificial intelligence related technologies such as deep neural networks can effectively improve the accuracy and efficiency of radiologists, clinicians and medical research in diagnosis and treatment. The basic principle of these artificial intelligence technologies is: select a specific sequence (such as T1, T2, DWIseries) of medical imaging data such as MRI and CT (following the DICOM data format specification) to train the artificial intelligence model, and then based on the trained model for other Medical imaging data for analysis and prediction. However, there is no mandatory requirement for the naming rules of medical image sequences in the DICOM protocol, and doctors and manufacturers are free to name the sequence descriptions. That is, t...

AI Technical Summary

Problems solved by technology

Artificial intelligence will analyze a specific sequence or a collection of specific sequences in DICOM image data when assisting in diagnosis. If the wrong sequence is selected, it will usually lead to incorrect results or no reference value.

[0003] Application practice has proved that using some single methods based on daily experience (such as string comparison based on sequence names, or judgment based on metadata tags in DICOM) to classify DICOM sequences often leads to artificial intelligence-assisted diagnosis generating false positive results

Images