Prediction of cancer-specific survival status of patients with early colorectal adenocarcinoma after endoscopic therapy based on machine learning algorithms
LI Zhihong
CAI Yingbin
WANG Yan
FAN Hua
Yiliminuer Ahemai
LI Zimei
Abstract：Objective:To construct a cancer-specific survival status prediction model for patients with early colorectal adenocarcinoma after endoscopic treatment used machine learning algorithms.Methods:Based on SEER database,the data of 1 786 patients with early colorectal adenocarcinoma after endoscopic treatment were obtained,and the information included age,sex,race,cancer primary site,degree of cancer cell differentiation,pathological type of cancer tissue,radiotherapy,chemotherapy,tumor size,pathological condition,and marital status were extracted.After univariate Logistic regression and multivariate Logistic regression analysis,independent influencing factors of survival prognosis of patients with early colorectal adenocarcinoma after endoscopic treatment were determined.The patients were divided into training set and test set at a ratio of 8∶2.the factors with statistical differences in regression analysis were substituted into Logistic regression,random forest,extreme gradient boosting,support vector machine,decision tree,gradient boosting decision tree which were constructed by machine learning algorithm.To interpret results based on optimal machine learning models.Results:The results of multivariate Logistic regression showed that age,cancer primary site,degree of cancer cell differentiation,tumor size,pathological condition,and marital status were independent influencing factors of survival prognosis of patients with early colorectal adenocarcinoma after endoscopic treatment(P<0.05).The area under the curve of receiver operator characteristic of random forest in the training set and test set were 0.876 and 0.858,respectively.And the F1 score were 0.791 and 0.739,respectively.The interpretability analysis of the model based on random forest showed that age,marital status and tumor size were more important,while higher age,larger tumor diameter,poor differentiation,and existence of submucosal infiltration were risk factors for death,and married were protective factors.Patients with the primary site of cancer in the right colon had poorer survival compared to those with the primary site of cancer in the left colon.Conclusions:The prognostic model constructed by machine learning for patients with colorectal cancer has good performance.It can provide accurate individualized prediction.
Keywords：machine learningearly colorectal canceradenocarcinomaendoscopic therapysurvival statepredictive modelsinfluencing factornursing
Publication Date：2024-07-25
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：9( 2459-2467 )