AUTHOR=Wang Di , Pan Bing , Huang Jin-Can , Chen Qing , Cui Song-Ping , Lang Ren , Lyu Shao-Cheng 

TITLE=Development and validation of machine learning models for predicting prognosis and guiding individualized postoperative chemotherapy: A real-world study of distal cholangiocarcinoma

JOURNAL=Frontiers in Oncology

VOLUME=Volume 13 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1106029

DOI=10.3389/fonc.2023.1106029

ISSN=2234-943X

ABSTRACT=Background: Distal cholangiocarcinoma (dCCA), originating from the common bile duct, is greatly associated with a dismal prognosis. In this study, we explored and compared several novel machine learning models which might lead to improvement in prediction accuracy and treatment options for patients with dCCA.
Methods: In this study, 169 patients with dCCA were recruited and randomly divided into the training cohort (n=118) and the validation cohort (n=51), and medical records were reviewed. Variables identified as independently associated with the primary outcome by following algorithms, LASSO regression, random survival forest (RSF), Univariate and Multivariate Cox regression analysis, were introduced to establish following different machine learning models and canonical regression model: Support Vector Machine (SVM), SurvivalTree, Coxboost, RSF, DeepSurv, and CoxPH. We measured and compared the performance of models with ROC, integrated Brier score (IBS) and C-index following cross-validation. The machine learning model with the best performance was screened out and compared with TNM Classification using ROC, IBS and C-index. Finally, Patients were stratified based on the model with the best performance to assess whether they benefited from postoperative chemotherapy.
Results: 5 variables, including tumour differentiation, T-stage, lymph node metastasis (LNM), AFR and CA19-9, were used to develop machine learning models. In the training cohort and the validation cohort, C-index of DeepSurv was better than other models. DeepSurv model had the highest mean AUC than other models. The IBS of DeepSurv model was lower than other models. Results of calibration chart and DCA also demonstrated DeepSurv had the satisfactory predictive performance. The performance of DeepSurv model was better than TNM Classification in C-index, mean AUC and IBS in the training cohort. Patients were stratified and divided into high- and low-risk group based on DeepSurv model. In the training cohort, patients with the high-risk group would not benefit from postoperative chemotherapy. In the low-risk group, patients receiving postoperative chemotherapy might have a better prognosis. 
Conclusions: DeepSurv model was good at predicting prognosis and risk stratification to guide treatment options. AFR level might be a potential prognostic factor for dCCA. For the low-risk group in DeepSurv model, patients might benefit from the post-operative chemotherapy.