Neuro-oncology · 2026
Gadolinium-free MR perfusion imaging based on generative adversarial network for primary intracranial tumor diagnosis: A multi-center study.
СтатьяМногоцентровое исследованиеНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Lin Guoqi, Ding Wangbin, Chen Yihua, Xiong Lukui, Hong Junhuan, He Pingping, Pan Yuwei, Tian Minxuan, Tao Jiahao, Guo Wei, Su Yan, Chen Ming, She Dejun
Аннотация
BACKGROUND: Accurate evaluation of tumor vascularity is essential in neuro-oncology, but conventional cerebral blood volume (CBV) mapping depends on gadolinium-based contrast agents. The aim was to develop a gadolinium-free model to synthesize CBV (CBVsyn) maps from non-contrast MRI (NC-MRI).
METHODS: This multicenter retrospective study analyzed 1227 MRI examinations in patients with primary intracranial tumors, comprising an internal cohort (train/validation/test 438/126/63) and 2 external cohorts (UPENN-GBM Cohort 464; UCSD-PTGBM cohort 136). A tumor-aware generative adversarial network (TA-GAN) was implemented to synthesize CBVsyn maps from various NC-MRI combinations. Image quality was evaluated using structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and a 4-point Likert scale. The correlation between pathologic markers and CBV parameters was conducted. Diagnostic performance was compared between CBVsyn and DSC-PWI-derived CBV (CBVreal) maps. The incremental diagnostic value of CBVsyn was evaluated.
RESULTS: TA-GAN utilizing T2-weighted and diffusion-weighted imaging achieved best performance in CBVsyn synthesis (PSNRGlobal=25.52 ± 1.33dB, SSIMGlobal=67.53 ± 5.88; both P < .05), with 95.34% maps being diagnostically acceptable. CBVsyn parameters showed positive correlation with CD105 markers (r = 0.49, P = .03). CBVsyn maps demonstrated comparable AUC with CBVreal maps for intracranial tumors grading (0.61 vs. 0.66, P = .51), pediatric glioma (0.60 vs. 0.57; P = .57), adult glioma (0.68 vs. 0.73; P = .56), and IDH1 genotype (0.66 vs. 0.81; P = .03). NC-MRI with CBVsyn assistance improved diagnostic accuracy (84.48% vs.75.86%) and consistency (κ = 0.83 vs. 0.62) for glioma grading compared with NC-MRI alone.
CONCLUSION: As a gadolinium-free alternative, CBVsyn maps synthesized by TA-GAN correlate with pathologic findings and demonstrate promise for characterizing primary intracranial tumors.
METHODS: This multicenter retrospective study analyzed 1227 MRI examinations in patients with primary intracranial tumors, comprising an internal cohort (train/validation/test 438/126/63) and 2 external cohorts (UPENN-GBM Cohort 464; UCSD-PTGBM cohort 136). A tumor-aware generative adversarial network (TA-GAN) was implemented to synthesize CBVsyn maps from various NC-MRI combinations. Image quality was evaluated using structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and a 4-point Likert scale. The correlation between pathologic markers and CBV parameters was conducted. Diagnostic performance was compared between CBVsyn and DSC-PWI-derived CBV (CBVreal) maps. The incremental diagnostic value of CBVsyn was evaluated.
RESULTS: TA-GAN utilizing T2-weighted and diffusion-weighted imaging achieved best performance in CBVsyn synthesis (PSNRGlobal=25.52 ± 1.33dB, SSIMGlobal=67.53 ± 5.88; both P < .05), with 95.34% maps being diagnostically acceptable. CBVsyn parameters showed positive correlation with CD105 markers (r = 0.49, P = .03). CBVsyn maps demonstrated comparable AUC with CBVreal maps for intracranial tumors grading (0.61 vs. 0.66, P = .51), pediatric glioma (0.60 vs. 0.57; P = .57), adult glioma (0.68 vs. 0.73; P = .56), and IDH1 genotype (0.66 vs. 0.81; P = .03). NC-MRI with CBVsyn assistance improved diagnostic accuracy (84.48% vs.75.86%) and consistency (κ = 0.83 vs. 0.62) for glioma grading compared with NC-MRI alone.
CONCLUSION: As a gadolinium-free alternative, CBVsyn maps synthesized by TA-GAN correlate with pathologic findings and demonstrate promise for characterizing primary intracranial tumors.