Feihong-cc / MAda-SR

Modal Adaptive Super-Resolution for MR and CT Scans Reconstruction via Continual Learning

MAda-SR

Modal Adaptive Super-Resolution for MR and CT Scans Reconstruction via Continual Learning

This repository is for MAda-SR introduced in the following paper. The code is built on HAN (PyTorch) and tested on Ubuntu 16.04/18.04 environment (Python3.9, PyTorch_1.12.0, CUDA11.7) with GeForce RTX3090 GPUs.

Contents

1. Introduction
2. Train
3. Test
4. Acknowledgements

Introduction

We proposed a multi-modal adaptive super-resolution algorithm for reconstructing CT and MRI scans, named MAda-SR, which improves the traditional Adam optimizer into an adaptive optimizer in terms of parameter updates and optimization strategies.

Begin to train

Begin to Test

Single task Test Cmd

python main.py --mode mhan --data_train Medical --data_test Medical --lml icarl -- reg_lambda 0.01 --scale 4 --pre_train ../experiment1/FFMx4_icarl/task_1_PD/model/model_latest.pt --save my_test --save_results

Multiple task Test Cmd

python multimain.py --model mhan --data_train Medical --data_test Medical --lml icarl -- reg_lambda 0.01 --scale 4 --pre_train ../experiment1/FFMx4_icarl/task_1_PD/model/model_latest.pt --save my_test --save_results

Acknowledgements

This code is built on HAN. We thank the authors for sharing their codes of HAN PyTorch version.