Preoperative prediction of lymph node metastasis using deep learning-based features

作     者：Renee Cattell Jia Ying Lan Lei Jie Ding Shenglan Chen Mario Serrano Sosa Chuan Huang 

作者机构：Department of Biomedical EngineeringStony Brook UniversityNY 11794 Stony BrookUSA Department of Radiation OncologyRenaissance School of MedicineStony Brook UniversityStony BrookNY 11794USA Program in Public HealthStony Brook MedicineStony BrookNY 11794USA Department of MedicineNorthside Hospital GwinnettGA 30046 LawrencevilleUSA Department of Radiation OncologyMedical College of WisconsinMilwaukeeWI 53226USA Institute of High Energy PhysicsChinese Academy of SciencesBeijing 100049China Department of RadiologyRenaissance School of MedicineStony Brook UniversityStony BrookNY 11794US 

出 版 物：《Visual Computing for Industry,Biomedicine,and Art》 (工医艺的可视计算（英文）)

年 卷 期：2022年第5卷第1期

页      面：88-98页

核心收录：

学科分类：0710[理学-生物学] 1002[医学-临床医学] 1001[医学-基础医学(可授医学、理学学位)] 0837[工学-安全科学与工程] 100214[医学-肿瘤学] 0714[理学-统计学（可授理学、经济学学位）] 0701[理学-数学] 10[医学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：This work was supported in part by National Cancer Institute,No.R03CA223052 Walk-for-Beauty Foundation and Baldwin Carol M.Baldwin Breast Cancer Research Awards。 

主　　题：Deep learning Radiomics Prediction model Lymph node metastasis Breast cancer 

摘      要：Lymph node involvement increases the risk of breast cancer recurrence.An accurate non-invasive assessment of nodal involvement is valuable in cancer staging,surgical risk,and cost savings.Radiomics has been proposed to pre-operatively predict sentinel lymph node(SLN)status;however,radiomic models are known to be sensitive to acquisition parameters.The purpose of this study was to develop a prediction model for preoperative prediction of SLN metastasis using deep learning-based(DLB)features and compare its predictive performance to state-of-the-art radiomics.Specifically,this study aimed to compare the generalizability of radiomics vs DLB features in an independent test set with dissimilar resolution.Dynamic contrast-enhancement images from 198 patients(67 positive SLNs)were used in this study.Of these subjects,163 had an in-plane resolution of 0.7×0.7 mm^(2),which were randomly divided into a training set(approximately 67%)and a validation set(approximately 33%).The remaining 35 subjects with a different in-plane resolution(0.78×0.78 mm^(2))were treated as independent testing set for generalizability.Two methods were employed:(1)conventional radiomics(CR),and(2)DLB features which replaced hand-curated features with pre-trained VGG-16 features.The threshold determined using the training set was applied to the independent validation and testing dataset.Same feature reduction,feature selection,model creation procedures were used for both approaches.In the validation set(same resolution as training),the DLB model outperformed the CR model(accuracy 83%vs 80%).Furthermore,in the independent testing set of the dissimilar resolution,the DLB model performed markedly better than the CR model(accuracy 77%vs 71%).The predictive performance of the DLB model outperformed the CR model for this task.More interestingly,these improvements were seen particularly in the independent testing set of dissimilar resolution.This could indicate that DLB features can ultimately result in a more generalizable model.