Pulmonary nodule image classification method when uncertain data is contained in data set

Abstract

The invention relates to the technical field of computer vision, and provides a pulmonary nodule image classification method when uncertain data is contained in a data set. The method comprises the following steps: firstly, collecting a pulmonary nodule CT image set, determining the category of the image through a majority voting principle by utilizing an expert voting method, and preprocessing toobtain a pulmonary nodule CT image data set; then, based on a knowledge distillation method, constructing a pulmonary nodule image classification model comprising a teacher model and a student model;next, obtaining a determined tag data set, training a teacher model on the determined tag data set, and calculating a soft tag on the pulmonary nodule CT image data set; then, training a student model on the data set combining the hard label and the soft label; and finally, inputting the preprocessed CT image to be classified into the trained lung nodule image classification model to obtain the category of the lung nodule image classification model. According to the method, the uncertain label data in the data set can be effectively utilized, the accuracy and efficiency of pulmonary nodule diagnosis are improved, and the usability and robustness are high.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a pulmonary nodule image classification method when uncertain data is contained in a data set. Background technique [0002] Lung cancer is one of the deadliest diseases in the world, accounting for approximately 26% of all cancer cases in 2017. Despite recent advances in diagnosis and treatment, the five-year cure rate for lung cancer is only 18%. It is worth noting that this rate can be greatly increased if patients are diagnosed early and properly treated. Low-dose computed tomography (CT) has been widely used in lung cancer detection. Compared with other imaging techniques, CT can display low-contrast nodules, which has more advantages. Low-dose CT scans can reduce lung cancer deaths by 20 percent, according to the National Lung Screening report. The traditional diagnosis of pulmonary nodules mainly relies on the observation of doctors and experts, so there are som...

AI Technical Summary

Problems solved by technology

The traditional diagnosis of pulmonary nodules mainly relies on the observation of doctors and experts, so there are some disadvantages: the diagnostic results are subjective; the workload is heavy and time-consuming; human vision is limited; doctors in remote areas cannot be popularized

Because of the professionalism and particularity of medical image labeling, as well as moral and legal restrictions, it is difficult to obtain a large amount of training data

In general public datasets include the following problems: (i) the dataset is small in size and the class distribution is unbalanced; (ii) it contains noise and uncertain labels

Taking the LIDC-IDRI dataset as an example, lung CT images are marked by multiple experts, each expert marks a part of the data, and each data is also marked by multiple experts, although the workload is reduced and relatively good results can be given. Labeling results, but different experts have different levels, which will produce a lot of inconsistent labels; secondly, many of the data are judged as uncertain types of data and unlabeled data. The traditional processing method is to directly convert uncertain types of data into Discard, but this approach will lose a lot of valuable information

Images