Quantitative study of the changes in brain white matter before and after radiotherapy by applying multi-sequence MR radiomics

Abstract

Abstract Purpose To analyse the changes in brain white matter before and after radiotherapy (RT) by applying multisequence MR radiomics features and to establish a relationship between the changes in radiomics features and radiation dose. Methods Eighty-eight patients with brain tumours who had undergone RT were selected in this study, and MR images (T1, T1+C, T2FLAIR, T2, DWI, and ASL) before and after RT were obtained. The brain white matter was delineated as an ROI under dose gradients of 0–5 Gy, 5–10 Gy, 10–15 Gy, 15–20 Gy, 20–30 Gy, 30–40 Gy, and 40–50 Gy. The radiomics features of each ROI were extracted, and the changes in radiomics features before and after RT for different sequences under different dose gradients were compared. Results At each dose gradient, statistically significant features of different MR sequences were mainly concentrated in three dose gradients, 5–10 Gy, 20–30 Gy, and 30–40 Gy. The T1+C sequence held the most features (66) under the 20–30 Gy dose gradient. There were 20 general features at dose gradients of 20–30 Gy, 30–40 Gy, and 40–50 Gy, and the changes in features first decreased and then increased following dose escalation. With dose gradients of 5–10 Gy and 10–15 Gy, only T1 and T2FLAIR had general features, and the rates of change were − 24.57% and − 29.32% for T1 and − 3.08% and − 10.87% for T2FLAIR, respectively. The changes showed an upward trend with increasing doses. For different MR sequences that were analysed under the same dose gradient, all sequences with 5–10 Gy, 20–30 Gy and 30–40 Gy had general features, except the T2FLAIR sequence, which was concentrated in the FirstOrder category feature, and the changes in features of T1 and T1+C were more significant than those of the other sequences. Conclusions MR radiomics features revealed microscopic changes in brain white matter before and after RT, although there was no constant dose-effect relationship for each feature. The changes in radiomics features in different sequences could reveal the radiation response of brain white matter to different doses.