CEST-KAN

This repository contains the code and demo data for CEST MRI data analysis using Kolmogorov-Arnold Networks (KAN) and Lorentzian-KAN (LKAN), as presented in the following paper:

Wang J, Cai P, Wang Z, Zhang H, Huang J*. CEST MRI data analysis using Kolmogorov-Arnold network (KAN) and Lorentzian-KAN (LKAN) models. Magnetic Resonance in Medicine 2025, 94(3), 1301-1317.

If you use the code, please consider citing our paper above. The original KAN can be found here, while the original efficient KAN can be found here.

All updated files are in the ‘New version’ folder

1. Training can be performed using train_cest_mlp.py, train_cest_kan.py and train_cest_lkan.py for MLP, KAN and LKAN from ‘New Version/Training.py’ folder, respectively.Testing can be performed using test_cest_mlp.py, test_cest_kan.py and test_cest_lkan.py for MLP, KAN and LKAN from ‘New Version/Testing.py’ folder, respectively.
2. Due to ethical issues, the original training data was not uploaded. Instead, 100,000 Z spectra generated based on real data (randomly selected and with noise added), together with their targets were uploaded for reproduction purposes. The training results are consistent with those shown in the paper (KAN and LKAN outperforms MLP).
3. The files (py/mat) starting with ‘Figure’ from 'New Version/' folder are used to generate the corresponding figures in the paper.

Reference: Wang J#, Cai P#, Wang Z, Zhang H, Huang J*. CEST-KAN: Kolmogorov-Arnold Networks for CEST MRI Data Analysis[J]. arXiv preprint arXiv:2406.16026, 2024.