Texture analysis of T2 weighted magnetic resonance imaging for the treatment effect of rectal cancer patients undergoing chemotherapy
Objective To explore the application of magnetic resonance T2 weighted image texture analysis in rectal cancer patients undergoing chemotherapy.Methods This study used 116 surgically treated rectal cancer patients who underwent surgical treatment and were admitted to Shandong Provincial Hospital from June 2019 to October 2022 as the research subjects,and they all underwent pathological examination and clinical examination to determine their diagnostic results.They were grouped according to response or not,namely non-complete response group(n=75)and complete response group(n=41).Magnetic resonance imaging was performed on patients before and after chemotherapy;Omni Kinetics software was applied to extract patient texture features,and energy parameters and variance,kurtosis,and entropy were calculated.ROC curve was used for texture analysis of magnetic resonance T2 weighted images to evaluate the pathological complete response of rectal cancer after chemotherapy.Results After treatment,the energy parameters(0.03±0.01)and kurtosis(3.71±0.50)of the non-complete response group were lower than those of the complete response group(0.05±0.01)and(4.84±0.67)(P<0.05).The variance(2311.00±587.00)and entropy(6.41±0.30)of the non-complete response group were higher than those of the complete response group(1533.00±313.00)and(5.42±0.42)(P<0.05).In the complete response group,the variance of magnetic resonance scanning texture analysis was negatively correlated with entropy(P<0.05),and positively correlated with kurtosis and energy(P<0.05).The specificity and sensitivity of energy parameters,methods,kurtosis and entropy of magnetic resonance scanning texture were not less than 85%,and the AUC was not less than 0.85.Conclusion By using MR T2 weighted image texture analysis,clinical practice can more accurately assess the pathological status of rectal cancer patients after chemotherapy treatment and thus better predict their complete response.
国家科技期刊平台
NSTL
万方数据
