Medical image noise data identification method based on artificial intelligence and related device

Abstract

The embodiment of the invention discloses a medical image noise data identification method based on artificial intelligence and a related device. For a medical image with a label, if the medical imagewith the label needs to identify whether noise data exists in the medical image, the medical image can be taken as to-be-identified label data and be identified through a first identification model.The first identification model has a good anti-noise capability. According to the identification result obtained by the first identification model, the reflected confidence coefficient can effectivelyexpress the possibility that the to-be-identified label data belongs to the noise data, so that the medical images with wrong label annotations are effectively screened out from the medical images with labels, and the determination accuracy of the noise data is improved.

Description

technical field [0001] This application relates to the field of data processing, in particular to a noise data identification method and related devices based on artificial intelligence for medical images. Background technique [0002] During network model training, labeled data can be used as training samples for supervised or semi-supervised training of the network model. Label data can be understood as data carrying a label, which is used to identify the specified content embodied and involved in the data. For example, in the label data about the identification of the lesion area in the medical image, the label of the medical image can reflect the data in the medical image. Whether there is a lesion area, or the lesion type of the lesion area, etc. [0003] However, not all data labels are accurate. If the label of a label data is actually incorrect, that is, the content identified by the label is not reflected in the data, or the data does not involve the content identi...

AI Technical Summary

Problems solved by technology

However, many noisy data do not belong to outliers under the statistical rules. For example, a medical image with a lesion identified by a label does not show a lesion, but shows a wound formed after surgery on the lesion area. Label data should belong to noise data, but it cannot be determined as an outlier from a statistical point of view

[0005] It can be seen that the accuracy of noise data screening methods for medical images in related technologies is not high, and it is difficult to meet the current noise data identification requirements

Images