2560 - Quantitative Study of the Changes in the Hippocampus after Radiotherapy Via Multisequence MR Radiomics

R. Liu^1,2, G. Gong¹, S. Du¹, K. Meng¹, W. Ruozheng³, and Y. Yin⁴; ¹Department of Radiation Oncology Physics and Technology, Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, shandong, China, ²Shandong First Medical University, Jinan, shandong, China, ³Tumor Hospital of Xinjiang Uygur Autonomous Region, wulumuqi, xinjiang, China, ⁴Department of Radiation Physics, Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, China

Purpose/Objective(s): Multisequence magnetic resonance (MR) radiomic features were used to analyze dynamic changes in the hippocampus after radiotherapy (RT), providing an objective basis for the early prediction of hippocampal radiation injury.

Materials/

Methods: The MR images of 75 patients with brain metastases (BMs) who received whole-brain radiotherapy (WBRT) were obtained before RT, after RT (26.22±14.05 days after the before RT scan), and at follow-up (393.45±296.33 days after the after RT scan). The imaging modalities included 3D T1-weighted imaging (T1WI), contrast-enhanced (CE) T1WI, 3D T2-weighted imaging (T2WI), T2 FLAIR imaging, and diffusion weighted imaging (DWI). Radiomic features were then extracted from contours of the hippocampus on all the different sequences. Changes in hippocampal volume and radiomic features from the different sequences before and after RT and at follow-up were analyzed. The feature analysis screening procedure was as follows: (1) Group₁ features were defined as radiomic features that were significantly different among the before RT, after RT, and follow-up time points; (2) Group₂ features were those that were not significantly different between the before and after RT time points but were significantly different separately between the before and after RT time points and the follow-up time point.

Results: (1) Compared with that before RT (3.41±0.49 cm³), the average hippocampal volume after RT and at follow-up was 3.32±0.49 cm³ and 2.95±0.45 cm³, respectively, representing reductions of 1.68% and 12.51% (p<0.05). (2) Group₁ primarily consisted of the T1WI and T2WI sequence features, 8 of which were common between the sequences. The feature with the highest rate of change (ROC) in the T1WI sequence was glcm-Contrast (20.92%-45.84%), while that with the highest ROC in the T2WI sequence was glcm-ClusterTendency (22.26%-59.59%). (3) Group₂ mainly consisted of CE-T1WI and T2WI sequence features, 11 of which were shared between them. The feature with the highest ROCs in the CE-T1WI and T2WI sequences was glcm-ClusterProminence for both (211.73%-344.97%; 143.97%-425.45%, respectively).

Conclusion: Compared to the hippocampal volume, the radiomic features from multisequence MR images could more intuitively reflect the dynamic microscopic changes in the hippocampus before and after RT (even during follow-up), which could provide an objective basis for tracking hippocampal changes and early prediction of injury after WBRT.