进展期胃癌与胃淋巴瘤的CT鉴别诊断及其判别模型的初步建立

Preliminary establishment of CT-based model for differentiation between advanced gastric cancer and gastric lymphoma

目的:探讨CT在进展期胃癌(AGC)与胃淋巴瘤(GL)鉴别诊断中的价值,初步建立AGC与GL的判别模型。方法:回顾性分析经病理证实的50例AGC及23例GL患者的CT资料,总结AGC与GL的CT鉴别诊断要点,并采用逐步判别分析及分批筛选变量的方法,初步建立两者的CT判别模型。结果:①AGC最常累及胃窦,GL最常累及胃体及胃窦;累及多部位在两组间差异无统计学意义(P=0.193)。②弥漫性胃壁增厚和节段性胃壁增厚在两组间的发生率差异无统计学意义,局限性胃壁增厚多见于AGC患者,两组间差异有统计学意义(P=0.036)。③GL组胃壁厚度大于AGC组,两组间差异有统计学意义(P=0.001)。胃壁厚度鉴别AGC与GL的ROC曲线下面积为0.754(P=0.001),以14.5 mm为阈值鉴别诊断AGC与GL的敏感度和特异度分别为82.6%、56.0%,Youdern指数为0.386。④形态固定、胃腔局限性狭窄多见于AGC组,两组间差异有统计学意义(P=0.008)。⑤AGC组多表现为明显强化,GL组多表现为轻度/中度强化,两组间差异有统计学意义(P=0.000);AGC组多见不均匀强化,GL组多见均匀强化,两组间差异有统计学意义(P=0.004)。⑥AGC组多见白线征,GL组多见胃黏膜面细线样强化,两组间差异均有统计学意义(P值均=0.000)。⑦AGC组多见胃浆膜面脂肪间隙模糊,GL组多见浆膜面脂肪间隙清晰,两组间差异有统计学意义(P=0.000)。⑧AGC组与GL组均可见腹腔内淋巴结异常,腹主动脉下部旁淋巴结异常GL组相对多见于AGC组,但差异无统计学意义(P=0.074);淋巴结坏死只见于AGC患者(12%),GL患者未发现淋巴结坏死。采用逐步判别分析法(分批筛选)筛选出胃壁厚度、胃腔形态、强化程度、黏膜面细线样强化等4个最有效的指标,初步建立AGC与GL的CT判别模型,交互验证法验证结果显示判别函数对AGC与GL的总体正确判别率为91.8%。结论:多种CT征象在AGC与GL的鉴别诊断中具有重要价值,经逐步判别分析法(分批筛选)初步建立的AGC与GL的CT判别模型,具有临床实用价值。

Objective:To explore the value of CT for the differential diagnosis of advanced gastric carcinoma(AGC)and gastric lymphoma(GL)and establish discrimination analysis model.Methods:CT images of 50 AGC patients and 23 GL patients confirmed by pathology were retrospectively analyzed.Imaging features were compared and CT discrimination analysis model were basically established using stepwise discriminant analysis and screening variables in batches.Results:①AGC mostly located at gastric antrum,while GL mostly located at gastric body and antrum.There was no significant difference for multi-sites involvement(P=0.193).②Diffuse and segmental gastric wall thickening showed no difference.Focal gastric wall thickening was more common in AGC(P=0.036).③The gastric wall thickness of GL group was higher than that of AGC group(P=0.001)Using the cutoff of 14.5mm,it had area under ROC of 0.754(P=0.001),sensitivity of 82.6%,and specificity of 56.0%,and the Youdern index of 0.386.④Fixed shape,gastric cavity narrow was more common in AGC group(P=0.008).⑤Intense enhancement was more common in the AGC group than GL group(P=0.000),inhomogeneous enhancement was more common in the AGC group than GL group(P=0.004).⑥In the AGC group,white line sign was more often seen(P=0.000),while in the GL group,fine line like enhancement of gastric mucosa was more common(P=0.000).⑦Blurred fat space on the serosal surface was more common in AGC group while clear fat space on the serosal surface in GL group(P=0.000).⑧there was no statistical significance for para-aortic lymphadenopathy(P=0.074).Internal necrosis was only found in AGC patients(12%).⑨Stepwise discriminant analysis(batch screening)screened out 4 most effective variables:gastric wall thickness,shape,enhancement degree,and fine line enhancement of mucosal surface.The CT-based discriminant analysis model was basically established,and the total correct discriminant rate of AGC and GL by interactive verification method was 91.8%.Conclusion:CT-based discriminant analysis model using stepwise discriminant analysis and screening variables in batches is valuable for the differential diagnosis between AGC and GL.