Image reconstruction method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and provides an image reconstruction method and device, electronic equipment and a storage medium, and the method comprises the steps: determining a to-be-reconstructed neural image, inputting the neural image into an image reconstruction model, and obtaining a reconstruction result of the neural image. An adopted image reconstruction model is obtained by training a target pre-training model through a first sample nerve image and a corresponding sample reconstruction result; and the target pre-training model is obtained by training in three pre-training modes of contrast learning unsupervised pre-training, cross-modal image conversion supervised pre-training and image reconstruction unsupervised pre-training, so that the problem of over-fitting of the model is avoided, the performance and generalization of the model in the aspect of an image reconstruction task are greatly improved, and on the basis, the target pre-training model is obtained by training in the three pre-training modes of contrast learning unsupervised pre-training, cross-modal image conversion supervised pre-training and image reconstruction unsupervised pre-training. The input neural image is reconstructed by using the image reconstruction model, so that the accuracy of the reconstruction result can be greatly improved.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, and in particular, to an image reconstruction method, device, electronic device and storage medium. Background technique [0002] Image reconstruction task refers to repairing defective images with missing parts, artifacts, low resolution, etc. This task is usually realized by neural network models. [0003] At present, most image reconstruction models based on transfer learning are trained by a single pre-training task combined with a conventional training method of fine-tuning. However, for brain neuroimaging reconstruction tasks, the scale of datasets is usually small, with only dozens or hundreds of cases, which makes conventional training methods prone to overfitting problems, resulting in poor performance of image reconstruction models. Difference. SUMMARY OF THE INVENTION [0004] The present invention provides an image reconstruction method, device, electronic ...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides an image reconstruction method, device, electronic equipment and storage medium, which are used to solve the defect that the training method of the image reconstruction model of neuroimaging in the prior art is prone to over-fitting

Images