Journal
COMPUTERS IN BIOLOGY AND MEDICINE
Volume 134, Issue -, Pages -Publisher
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.compbiomed.2021.104504
Keywords
Multi-channel MRI; Image reconstruction; Generative adversarial networks; Transfer learning
Categories
Funding
- National Natural Science Foundation of China [61902338, 62001120]
- Shanghai Sailing Program [20YF1402400]
- European Research Council Innovative Medicines Initiative on Development of Therapeutics and Diagnostics Combatting Coronavirus Infections Award [H2020-JTI-IMI2 101005122]
- British Heart Foundation [PG/16/78/32402]
- AI for Health Imaging Award [H2020-SC1-FA-DTS-2019-1 952172]
- Hangzhou Economic and Technological Development Area Strategical Grant [Imperial Institute of Advanced Technology]
Ask authors/readers for more resources
This study explored novel applications based on transfer learning using parallel imaging combined with the GAN model (PI-GAN), showing that transfer learning on brain tumor datasets could remove artifacts and yield smoother brain edges. Transfer learning results for knee and liver datasets were superior to those of the PI-GAN model trained with fewer training samples, although the learning process was slower in knee datasets compared to brain tumor datasets. Transfer learning improved reconstruction performance in models with acceleration factors of 2 and 6, with the AF=2 model showing better results. Transfer learning with pre-trained models helped address inconsistencies between training and test datasets and facilitated generalization to unseen data.
Deep learning based generative adversarial networks (GAN) can effectively perform image reconstruction with under-sampled MR data. In general, a large number of training samples are required to improve the reconstruction performance of a certain model. However, in real clinical applications, it is difficult to obtain tens of thousands of raw patient data to train the model since saving k-space data is not in the routine clinical flow. Therefore, enhancing the generalizability of a network based on small samples is urgently needed. In this study, three novel applications were explored based on parallel imaging combined with the GAN model (PI-GAN) and transfer learning. The model was pre-trained with public Calgary brain images and then fine-tuned for use in (1) patients with tumors in our center; (2) different anatomies, including knee and liver; (3) different k-space sampling masks with acceleration factors (AFs) of 2 and 6. As for the brain tumor dataset, the transfer learning results could remove the artifacts found in PI-GAN and yield smoother brain edges. The transfer learning results for the knee and liver were superior to those of the PI-GAN model trained with its own dataset using a smaller number of training cases. However, the learning procedure converged more slowly in the knee datasets compared to the learning in the brain tumor datasets. The reconstruction performance was improved by transfer learning both in the models with AFs of 2 and 6. Of these two models, the one with AF = 2 showed better results. The results also showed that transfer learning with the pre-trained model could solve the problem of inconsistency between the training and test datasets and facilitate generalization to unseen data.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available