Correlation Between CT Low-density Volume And Necrosis Tumor Ratio In Colorectal Cancer

PurposeAfter years of exploration and practice,significant progress has been made in the early diagnosis,treatment and prognosis prediction of colorectal cancer.However,the incidence of colorectal cancer is still on the rise,prompting people to continue to study the factors affecting the prognosis of colorectal cancer.Aiming at the clinical challenge that it is difficult to accurately predict the prognosis of colorectal cancer,this study intends to:firstly,perform artificial intelligence to automatically segment hematoxylin and eosin(HE)stained whole-slice images(WSIs)to obtain an automatic necrosis tumor ratio,and to explore the prognostic value of necrosis tumor ratio automatically identified by artificial intelligence for colorectal cancer.Secondly,to explore the correlation between computed tomography(CT)low-density volume obtained in portal vein phase and automated necrosis tumor ratio,and further investigated the prognostic value of CT low-density volume.Materials and methodsThe clinicopathologic characteristics and surgical specimens of UICC TNM I-III stage colorectal cancer patients from Shanxi Cancer Hospital and The Sixth Affiliated Hospital of Sun Yat-Sen University were retrospectively collected,as well as preoperative CT-enhanced images of colorectal cancer patients from Shanxi Cancer Hospital.The prespecified endpoints were disease-free survival(DFS)and overall survival(OS).In the first part of the study,HE-stained WSIs were segmented by U-Net to obtain 12 tissue types,including necrosis area and tumor area.Necrosis tumor ratio was defined as the sum of necrosis area divided by the sum of tumor epithelial area plus necrosis area in all available sections.In the second part of the study,we obtained preoperative CT-enhanced images in the portal vein phase that were available in Shanxi Cancer Hospital.According to the CT value distribution of voxel points in the tumor area,CT value less than 66th percentile was defined as CT value-low.Correlation analysis was performed between CT low-density volume and necrosis tumor ratio and CT low-density volume in predicting the prognosis of colorectal cancer was analyzed.In the above study,the difference of survival rate between different groups was determined by Kaplan-Meier curve,and the P value was calculated by log-rank test.Multivariate Cox proportional hazard model was used to analyze the correlation between various factors and OS and DFS.ResultsPart one of the study:Patients with a low necrosis tumor ratio showed a significant better prognosis.The unadjusted hazard ratio(HR)for necrosis tumor ratio-high vs necrosis tumor ratio-low was 1.72(95%confidence interval[CI]1.19-2.49,P=0.004).In the validation cohort,these findings were confirmed(1.98[1.22-3.23],0.006).In addition,this study found that patients classified as necrosis tumor ratio-high could benefit from adjuvant chemotherapy in patients with stage II colorectal cancer(P=0.047).In necrosis tumor ratio-low group,there was no difference in OS and DFS between the surgery alone group and adjuvant chemotherapy group(P>0.05).Part two of the study:Based on the percentiles of 25%,50%and 75%,CT lowdensity volume was divided into low,middle and high groups.The average necrosis tumor ratio of the CT low-density volume-high was higher than that of the CT lowdensity volume-low group(P<0.0001).In addition,CT low-density volume was significantly correlated with DFS(P>0.05).The 3-year DFS rate decreased from 89.2%in the CT low-density volume-low group to 76.2%in the CT low-density volume-high group.The unadjusted HR for the CT low-density volume-high vs low was 2.02(95%CI 1.16-3.52,P=0.012).ConclusionNecrosis tumor ratio based on fully automatic quantification of digital pathology could predict the prognosis of patients with colorectal cancer and the efficiency of adjuvant chemotherapy in stage Ⅱ patients.CT low-density volume in the tumor area quantified in the portal phase of preoperative CT was strongly correlated with necrosis tumor ratio,which could stratify the prognostic risk of colorectal cancer.Through the quantitative analysis of images,pathological data and clinical data,we can assist the personalized diagnosis and treatment and accurate prediction of diseases.