基于一体化^(18)F-FDG PET/MRI深度学习模型评估脑梗死周围脑代谢

Deep learning model based on integrated ^(18)F-FDG PET/MRI for evaluating cerebral metabolism around cerebral infarction

目的探讨基于一体化^(18)F-FDG PET/MRI深度学习(DL)模型评估脑梗死周围脑代谢状态的价值。方法回顾性收集46例单侧颈内动脉(ICA)或大脑中动脉(MCA)狭窄闭塞致脑梗死患者,基于头部一体化^(18)F-FDG PET/MRI以DL模型自动分割脑梗死区域,应用不对称指数(AI)评估分割出的脑梗死区域中额叶、颞叶、顶叶、枕叶及大脑半球代谢减低区域体积及其占比,并分析其与美国国立卫生研究院卒中量表(NIHSS)神经功能评分的相关性。结果46例患侧颞叶、顶叶及大脑半球中的代谢减低区域体积分别为(41.35±10.52)ml、(65.58±14.82)ml和(178.89±34.23)ml,均与NIHSS评分呈正相关(rs=0.359、0.343、0.362,P均<0.05);患侧额叶、颞叶、顶叶及大脑半球中的代谢减低区域体积占比分别为(45.68±10.35)%、(42.32±10.19)%、(45.05±9.44)%及(44.11±8.63)%,均与NIHSS评分呈正相关(rs=0.344、0.340、0.439、0.393,P均<0.05)。结论基于一体化^(18)F-FDG PET/MRI DL模型评估脑梗死周围脑代谢状态具有重要临床价值。

Objective To investigate the value of deep learning(DL)model based on integrated ^(18)F-FDG PET/MRI for evaluating cerebral metabolic status around cerebral infarction.Methods A total of 46 patients with cerebral infarction caused by unilateral internal carotid artery(ICA)or middle cerebral artery(MCA)steno-occlusion were retrospectively collected.Based on integrated ^(18)F-FDG PET/MRI,DL model was used to automatically segment cerebral infarction area.Asymmetry index(AI)was used to evaluate the volume of reduced metabolic areas in the segmented affected frontal lobe,temporal lobe,parietal lobe,occipital lobe and cerebral hemisphere of cerebral infarction area as well as their proportions,while their correlations with National Institutes of Health stroke scale(NIHSS)score of neurological function were analyzed.Results Among 46 patients,the volume of decreased metabolism in the affected temporal lobe,parietal lobe and cerebral hemisphere was(41.35±10.52)ml,(65.58±14.82)ml and(178.89±34.23)ml,respectively,all positively correlated with NIHSS scores(rs=0.359,0.343,0.362,all P<0.05).The proportion of the reduced metabolic volume in the affected frontal lobe,temporal lobe,parietal lobe and cerebral hemisphere was(45.68±10.35)%,(42.32±10.19)%,(45.05±9.44)%and(44.11±8.63)%,respectively,all positively correlated with NIHSS scores(rs=0.344,0.340,0.439,0.393,all P<0.05).Conclusion DL model based on integrated ^(18)F-FDG PET/MRI was of important clinical value for evaluating cerebral metabolic state around cerebral infarction.