Scholarlink

<p>Survival Analysis is a statistical technique that aims in predicting the time of occurrence of an event. Survival analysis using machine learning (ML) plays a major role in clinical decision making. It acts as an important prognostic tool in deciding the cancer treatment options. Clinical data is often prone to have unlabelled data. Data labels are essential for machine learning models to predict the outcome. Manual labelling incurs cost and effort. To address this challenge, a self-supervised learning (SSL) based survival prediction is proposed. The proposed method adopts pretext learning as the SSL approach to handle the unlabelled data. The pretext learning learns the contextual embeddings from the unlabelled data by performing an auxiliary task. The SSL model incorporates survival aware pseudo labelling which enables the model to learn the labels from the data itself. The learned model is then used for survival classification and prediction. The proposed method is evaluated on breast cancer data from Surveillance Epidemiology End Results (SEER) program. The performance metrics Accuracy, Area Under the Curve (AUC), Recall, Precision and F1-Score are used for evaluating the classification models. Among the classifiers employed Support Vector Machine (SVM) performed well with an accuracy of 0.97, AUC of 0.99 precision of 0.97, recall of 0.97 and F1-Score of 0.97. The survival prediction models are evaluated using Concordance index(C-index). The survival model random survival forest (RSF) achieved the highest C-index of 0.97. The results show that the latent data outperforms the models learned using raw data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Auto-encoder Based Survival Prediction for Breast Cancer Patients Using Self-supervised Learning Approach

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月6日

查看全文
  • Deepa P,
  • Gunavathi C

摘要

Survival Analysis is a statistical technique that aims in predicting the time of occurrence of an event. Survival analysis using machine learning (ML) plays a major role in clinical decision making. It acts as an important prognostic tool in deciding the cancer treatment options. Clinical data is often prone to have unlabelled data. Data labels are essential for machine learning models to predict the outcome. Manual labelling incurs cost and effort. To address this challenge, a self-supervised learning (SSL) based survival prediction is proposed. The proposed method adopts pretext learning as the SSL approach to handle the unlabelled data. The pretext learning learns the contextual embeddings from the unlabelled data by performing an auxiliary task. The SSL model incorporates survival aware pseudo labelling which enables the model to learn the labels from the data itself. The learned model is then used for survival classification and prediction. The proposed method is evaluated on breast cancer data from Surveillance Epidemiology End Results (SEER) program. The performance metrics Accuracy, Area Under the Curve (AUC), Recall, Precision and F1-Score are used for evaluating the classification models. Among the classifiers employed Support Vector Machine (SVM) performed well with an accuracy of 0.97, AUC of 0.99 precision of 0.97, recall of 0.97 and F1-Score of 0.97. The survival prediction models are evaluated using Concordance index(C-index). The survival model random survival forest (RSF) achieved the highest C-index of 0.97. The results show that the latent data outperforms the models learned using raw data.