论文部分内容阅读
目的:探讨宝石能谱CT及机器学习算法在判断胃癌浆膜浸润中的价值。方法:回顾性分析在我院行宝石能谱CT双期GSI增强检查的胃癌患者24例,其中p T2 8例,p T3 4例,p T4 12例。12例患者(p T4)归为浆膜阳性组(组A);12例(T2和T3)归为浆膜阴性组(组B)。采用独立样本t检验或卡方检验比较两组患者的临床信息(如性别、年龄等)。此外,所有图像利用GE AW4.4工作站进行后处理,分别得出两组病灶双期能谱信息,随后采用SVM-RFE算法对两组能谱信息进行分析。结果:两组患者的临床信息中,肿瘤长径和短径在两组间有统计学差异(P均<0.05)。SVM-RFE算法的准确率为87.5%-94.4%。SVM-RFE的输出结果为门脉期脂肪(钙)、门脉期尿酸(钙)、动脉期钙(碘)、门脉期水(钙)、门脉期碘(水)。结论:肿瘤大小和门脉期脂肪(钙)、门脉期尿酸(钙)、动脉期钙(碘)、门脉期水(钙)及门脉期碘(水)特征值可用于辅助判定胃癌是否浸润浆膜层。
Objective: To explore the value of gem energy spectrum CT and machine learning algorithm in determining gastric serosa infiltration. Methods: Twenty-four patients with gastric cancer who underwent GSI enhanced gem spectroscopy in our hospital were retrospectively analyzed. Of them, 8 were in pT2, 4 in pT3, and 12 in pT4. Twelve patients (pT4) were classified as serosal positive (group A); 12 (T2 and T3) were classified as serosal negative (group B). Independent samples t-test or chi-square test were used to compare the clinical information (such as gender, age, etc.) between the two groups. In addition, all the images were processed by GE AW4.4 workstation to obtain two sets of lesion double-phase energy spectrum information respectively, and then the two groups of energy spectrum information were analyzed by SVM-RFE algorithm. Results: In the two groups of patients’ clinical information, there was a significant difference in the long diameter and the short diameter between the two groups (all P <0.05). The accuracy of SVM-RFE algorithm is 87.5% -94.4%. The output of SVM-RFE was portal venous fat (calcium), portal uric acid (calcium), arterial calcium (iodine), portal phase water (calcium) and portal phase iodine (water). Conclusions: The features of tumor size and portal venous fat (calcium), portal uric acid (calcium), arterial calcium (iodine), portal water (calcium) and portal venous iodine (water) Whether to immerse the serosa layer.