关键词: 直肠肿瘤, 肿瘤浸润, 图像处理，计算机辅助, 弥散加权成像

Abstract:

Objective To explore the diagnostic value of preoperative diffusion weighted imaging （DWI） histogram parameters in the depth of invasion of early rectal cancer. Methods A total of 180 patients with early rectal cancer admitted to 904th Hospital of Joint Logistics Support Force of Chinese People's Liberation Army from August 2020 to August 2024 were selected as the study objects. Patients were divided into intramucosal cancer group （n=102） and submucosal cancer group （n=78） according to the depth of tumor invasion. The general data of the two groups were compared. The intraclass correlation coefficient （ICC） was used to analyze the consistency of DWI histogram parameters extracted by the two radiologists， and the differences between the two groups were compared. Receiver operator characteristic （ROC） curve was used to analyze the predictive value of each parameter to the depth of tumor invasion. Multivariate logistic regression was used to analyze the independent influencing factors of invasion depth， and a predictive model was constructed. The ROC curve was drawn to analyze the predictive value of the model for tumor invasion depth， and the Hosmer-Lemeshow test was used to analyze the goodness of fit of the model. Results There were statistically significant differences in age （t=8.15， P<0.001）， maximum tumor diameter （χ²=29.29， P<0.001）， endoscopic type （χ²=20.96， P<0.001）， histological type （χ²=24.93， P<0.001） and differentiation degree （χ²=73.35， P<0.001） between intramucosal cancer group and submucosal cancer group. The mean， variance， skewness， kurtosis， the 1^st， 10^th， 50^th， 90^th， and 99^th percentiles of the histogram parameters of DWI had good consistency （all ICC>0.75）. There were statistically significant differences in the mean （t=5.69， P<0.001）， variance （t=9.75， P<0.001）， skewness （t=10.88， P<0.001）， kurtosis （t=10.06， P<0.001）， the 1^st percentile （t=3.43， P<0.001）， 10^th percentile （t=3.59， P<0.001）， 50^th percentile （t=9.97， P<0.001）， 90^th percentile （t=4.63， P<0.001）， and 99^th percentile （t=2.44， P=0.016） of the DWI histogram parameters between the intramucosal cancer group and the submucosal cancer group. ROC curve analysis results showed that mean [area under the curve （AUC）=0.77]， variance （AUC=0.88）， skewness （AUC=0.88）， kurtosis （AUC=0.78）， 50^th percentile （AUC=0.86） and 90^th percentile （AUC=0.82） had certain diagnostic value for submucous cancer. Multivariate analysis showed that age （OR=9.98， 95%CI： 1.10-90.70， P=0.041）， maximum tumor diameter （OR=7.36， 95%CI： 1.08-50.23， P=0.042）， and differentiation degree （OR=19.88， 95%CI： 1.21-327.92， P=0.037）， variance （OR=16.24， 95%CI： 2.26-116.68， P=0.006）， skewness （OR=21.13， 95%CI： 2.80-59.61， P=0.003）， 1^st percentile （OR=9.78， 95%CI： 1.17-81.76， P=0.035） were independent factors in predicting tumor invasion depth in patients with early rectal cancer. The predictive model based on the above indicators was logit（P）=1.51+2.30×age+2.00×maximum tumor diameter+2.99×differentiation degree+2.79×variance+3.05×skewness+ 2.28×the 1^st percentile. ROC curve analysis showed that the predictive model had an AUC of 0.97 （95%CI： 0.95-0.99） for judging the occurrence of submucosal cancer in patients with early rectal cancer， the sensitivity was 0.95， and the specificity was 0.88. The Hosmer-Lemeshow test results showed that the goodness of fit of the model was ideal （P=0.823）. Conclusions Age， maximum tumor diameter， differentiation degree， variance， skewness， and the 1^st percentile are independent factors in predicting tumor invasion depth in patients with early rectal cancer. The predictive model constructed based on these factors can effectively predict the risk of submucosal cancer in patients with early rectal cancer.

Key words: Rectal neoplasms, Neoplasm invasiveness, Image processing， computer-assisted, Diffusion weighted imaging