Multisequence MRI improves deep learning radiotherapy dose prediction in glioblastoma versus CT alone

Adding multisequence MRI to CT improved deep learning dose prediction for glioblastoma radiotherapy: mean absolute error was 5.99 Gy vs 6.92 Gy (p=0.006), with better spatial agreement (SSIM 0.599 vs 0.564) and high-dose region estimation.

• The retrospective cohort included 140 patients treated with 60 Gy in 30 fractions; training/validation/test split was 100/20/20.
• D2 error fell from 3.51 ± 1.61 Gy to 1.53 ± 1.53 Gy, and PSNR rose from 16.84 ± 2.17 dB to 18.01 ± 1.92 dB (p=0.009).
• The authors state further multicentre validation is required.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.