Scholarlink

Gallbladder cancer (GBC) is highly aggressive and clinically unpredictable, making the prognosis and treatments quite complex. This study aims to investigate whether an existing federated learning (FL) framework, together with the FedProx algorithm, can improve predictive analysis for GBC patients. We used CT scan data and integrated radiomic features with clinicopathological data to create an accurate prognostic model. The proposed FL framework enabled by the FedProx algorithm consists of a simply integrated radiomic-clinical dataset providing information accuracy, precision, recall, and F1-score based on the training and testing dataset of the federated K-nearest neighbors’ (FedKNN) model. Our results show that the FedProx algorithm, through the incorporation of FedKNN, can perform and attain an accuracy of approximately 92.86% on the training set and 89.66% on the test set, demonstrating the model’s ability to predict the outcomes of gallbladder cancers and its applicability in combining various data sources for enhanced predictive capabilities in oncology.

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

Enhancing Gall Bladder Cancer Detection by Privacy-Preserving Federated Learning: A Comprehensive CSV-Based Classification Approach

  • 章节

  • 发表日期：

    2026年

查看全文
  • Bidita Sarkar Diba,
  • Nishat Tasnim,
  • Mehjabin Hossain,
  • Aloke Kumar Saha,
  • Md. Golam Kibriya,
  • Moshiur Rahman Tonmoy

摘要

Gallbladder cancer (GBC) is highly aggressive and clinically unpredictable, making the prognosis and treatments quite complex. This study aims to investigate whether an existing federated learning (FL) framework, together with the FedProx algorithm, can improve predictive analysis for GBC patients. We used CT scan data and integrated radiomic features with clinicopathological data to create an accurate prognostic model. The proposed FL framework enabled by the FedProx algorithm consists of a simply integrated radiomic-clinical dataset providing information accuracy, precision, recall, and F1-score based on the training and testing dataset of the federated K-nearest neighbors’ (FedKNN) model. Our results show that the FedProx algorithm, through the incorporation of FedKNN, can perform and attain an accuracy of approximately 92.86% on the training set and 89.66% on the test set, demonstrating the model’s ability to predict the outcomes of gallbladder cancers and its applicability in combining various data sources for enhanced predictive capabilities in oncology.