Abstract:

Objective To investigate the diagnostic value of ultrasound texture analysis in T stage of rectal cancer. Methods The preoperative ultrasound images of 31 patients with rectal cancer confirmed by surgery or pathology were selected， and the region of interest （ROI） was delineated by mazda software. The software automatically calculated the texture characteristic parameters of mean， kurtosis， skewness， contrast， entropy， variance sum， autocorrelation， energy and deficit distance， compared the differences of texture parameters of rectal cancer in different stages， and drew the subject's working characteristic curve （ROC） for the statistically significant parameters， compared the area under the curve （AUC）， and evaluated its diagnostic efficiency. Results It was statistically shown that kurtosis， skewness， contrast， entropy， variance sum and autocorrelation were all influencing factors of T stage of rectal cancer， and the skewness， entropy， variance sum and autocorrelation of advanced rectal cancer （pT3-4 stage） were higher than those of early rectal cancer （pT1-2 stage） （P < 0.05）， while the kurtosis and contrast were lower than those of early rectal cancer （P < 0.05）. Among them， entropy has a higher diagnostic efficiency， and the AUC value is 0.88. Conclusion Ultrasound texture analysis can provide a more objective imaging basis for T staging of rectal cancer， and the texture characteristics of advanced rectal cancer are more complex than those of early rectal cancer.

Key words: rectal cancer, ultrasound， texture analysis, T stage