Expansion and enhancement method and system suitable for deep learning training data and readable storage medium

Abstract

The invention discloses an expansion and enhancement method and system suitable for deep learning training data and a readable storage medium, and the method at least comprises the following steps: S1, obtaining original training data, and based on the ray energy spectrum of the original training data, classifying the original training data through a data classification module, classifying the original training data with the same characteristic energy spectrum Sk into the same class; s2, N pieces of first data are selected from the original training data with the same characteristic energy spectrum Sk through a first data training module based on a first random selection mode, and the N pieces of first data generate first new data according to a first linear superposition mode; s3, M pieces of second data are selected from the first new data through a second data training module on the basis of a second random selection mode, and the M pieces of second data generate second new data according to a second linear superposition mode; and S4, repeating the steps S2 and S3 to generate a training sample.

Description

technical field [0001] The present invention relates to the technical field of data enhancement, in particular to an expansion and enhancement method, system and readable storage medium suitable for deep learning training data. Background technique [0002] The biggest challenge in the application of neural networks in the medical field is the lack of a large amount of labeled data. Using large-scale training sets that cover all situations as much as possible can improve the generalizability of the model and prevent the model from The performance is insufficient when encountering new data, but because unlike natural image datasets, medical image datasets are generally small, mainly because obtaining anonymous and accurately annotated medical image data is a laborious and difficult task. Similarly, in radiation therapy dose calculation, it is also difficult to obtain all dose distributions with different intensities and shapes. Therefore, the use of effective data enhancemen...

AI Technical Summary

Problems solved by technology

[0004] The existing technology does not focus on the principles and characteristics of dosimetry, and effectively adds useful new types of data for dose calculation, and may even generate erroneous new data, for example: 1) Cutting, because the dose at a certain point of dose calculation includes surrounding scattered rays Contribution, so the cutting operation may cause the loss of part of the scattering information

2) Scaling, particle interaction, and migration are related to the actual physical distance, such as the inverse square law of distance. Simple scaling without modifying the corresponding values ​​will lead to incorrect information reflected in this part

Images