Application of Machine Learning Algorithms to Predict Lymph Node Metastasis in Early Gastric Cancer

Abstract

ObjectiveThis study aimed to establish the best early gastric cancer lymph node metastasis (LNM) prediction model through machine learning (ML) to better guide clinical diagnosis and treatment decisions.MethodsWe screened gastric cancer patients with T1a and T1b stages from 2010 to 2015 in the Surveillance, Epidemiology and End Results (SEER) database and collected the clinicopathological data of patients with early gastric cancer who were treated with surgery at the Second Affiliated Hospital of Nanchang University from January 2014 to December 2016. At the same time, we applied 7 ML algorithms—the generalized linear model (GLM), RPART, random forest (RF), gradient boosting machine (GBM), support vector machine (SVM), regularized dual averaging (RDA), and the neural network (NNET)—and combined them with patient pathological information to develop the best prediction model for early gastric cancer lymph node metastasis. Among the SEER set, 80% were randomly selected to train the models, while the remaining 20% were used for testing. The data from the Second Affiliated Hospital were considered as the external verification set. Finally, we used the AUROC, F1-score value, sensitivity, and specificity to evaluate the performance of the model.ResultsThe tumour size, tumour grade, and depth of tumour invasion were independent risk factors for early gastric cancer LNM. Comprehensive comparison of the prediction model performance of the training set and test set showed that the RDA model had the best prediction performance (F1-score = 0.773; AUROC = 0.742). The AUROC of the external validation set was 0.73.ConclusionsTumour size, tumour grade, and depth of tumour invasion were independent risk factors for early gastric cancer LNM. ML predicted LNM risk more accurately, and the RDA model had the best predictive performance and could better guide clinical diagnosis and treatment decisions.

Machine learning for lymph node metastasis prediction of in patients with gastric cancer: A systematic review and meta-analysis

“…Larger tumor typically indicates a higher risk of LNM ( 86 – 88 ), which might be attributed to easiness of invasion for larger tumor to the surrounding tissues. Depth of tumor invasion was found to be a strong predictor in 32 models ( 32 , 34 – 37 , 39 , 41 , 42 , 46 , 49 , 50 , 52 , 54 – 56 , 62 , 69 – 71 ), which is in consistence with substantial evidence supporting its use as a predictor of LNM ( 36 , 63 , 89 – 91 ). There were 20 models considered histology differentiation as a vital factor for predicting LNM ( 34 – 37 , 41 , 42 , 48 – 55 , 61 – 63 ).…”

“…There were 35 studies (85.4%) that were conducted in China ( 32 , 33 , 35 – 43 , 45 – 49 , 51 – 54 , 56 – 63 , 65 – 69 , 71 , 72 ), 5 (12.2%) in Korea ( 34 , 44 , 50 , 55 , 64 ) and 1 (2.4%) in Germany ( 70 ), with the publication date ranged from 2004 to 2022. The number of studies using ML for LNM prediction has gradually increased since 2018 ( Figure 2 ).…”

“…Eight studies lacked external validation in model development ( 35 , 50 , 51 , 53 , 59 , 63 , 65 , 70 ). Concern regarding ‘overall applicability’ was rated as low in 15 studies (36.59%) ( 32 , 34 , 37 , 38 , 40 , 42 , 44 , 54 – 56 , 60 , 62 , 66 , 67 , 69 ), high in 16 studies (39.02%) ( 33 , 35 , 41 , 43 , 47 , 50 , 53 , 58 , 59 , 61 , 63 , 65 , 68 , 70 , 71 ) and unclear in the remaining 10 (24.39%) ( 36 , 39 , 45 , 46 , 48 , 49 , 52 , 57 , 64 , 72 ). Risk of bias and applicability assessment were shown in Table 2 .…”

“…Of the 41 included studies, 21 were early-gastric GC (EGC) (T1) studies ( 32 , 34 , 35 , 37 , 39 , 42 , 46 , 48 – 53 , 55 , 56 , 60 , 64 , 65 , 69 , 71 , 72 ) and 3 were advanced GC (T2-4) studies ( 43 , 45 , 67 ). In EGC, there was a pooled c-index of 0.832 [95%CI (0.804, 0.860)] ( Table 5 ; eFigure 13 ) and 0.795 [95%CI (0.755, 0.838)] ( eFigure 14 ) for the training and test groups, respectively.…”

“…There were 20 models considered histology differentiation as a vital factor for predicting LNM ( 34 – 37 , 41 , 42 , 48 – 55 , 61 – 63 ). Deeply infiltrated and poorly differentiated tumors might have sufficient nutritional support to facilitate its invasion to tissues, capillaries and lymphatic vessels, and thus to have the potential to grow and metastasize faster ( 69 ).…”