Scholarlink

<p>The early and accurate diagnosis of lung disease is still a big challenge in the clinic, especially with small and small-sized biomedical data. Plasmonic biosensors can be used for sensitive detection of variations in biomarkers associated with the progression of lung diseases, but signals from the biosensors are commonly noisy, nonlinear, and affected by intra-class variability, which makes the classification of the signals challenging. Besides, the healthcare data should be handled with care, which is crucial for its clinical deployment. In this study, we suggest a machine learning framework for the diagnosis of lung disease that is both interpretable and secured by blockchain using biosensor signals. The proposed framework involves preprocessing the biosensor data, clustering using the K-means algorithm, classification with the XGBoost model, interpretability analysis with SHAP, and data security through blockchain integration. Before classification, clustering was added to minimize intra-class variability, thereby enhancing the separation of patient cluster patterns. The integration of Blockchain with IPFS-based storage was integrated to provide a secure, tamper-proof, and privacy-preserving environment for handling healthcare information. The proposed model yielded 95.80% accuracy, 0.95 F1-score, and 0.98 ROC-AUC when trained on 90% of the data and tested on 10% of the data. Stratified 10-fold cross-validation (10-fold CV) had an accuracy of 94.12% and an ROC-AUC of 0.97. Noise perturbation analysis kept the accuracy high at 92.70% distortion for 10% distortion. The implementation of blockchain resulted in 100% data integrity and low latency (1.8&#xa0;s). The proposed lung-diagnosis scheme using plasmonic biosensor data is accurate, interpretable, computationally efficient, and secure.</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

Blockchain-secured and generalizable lung disease diagnosis from plasmonic biosensor signals using cluster-guided XGBoost with ablation and efficiency analysis

  • 期刊论文

  • 发表日期：

    2026年7月21日

查看全文
  • Veerapandian Saranyah,
  • Natesan Suganthi

摘要

The early and accurate diagnosis of lung disease is still a big challenge in the clinic, especially with small and small-sized biomedical data. Plasmonic biosensors can be used for sensitive detection of variations in biomarkers associated with the progression of lung diseases, but signals from the biosensors are commonly noisy, nonlinear, and affected by intra-class variability, which makes the classification of the signals challenging. Besides, the healthcare data should be handled with care, which is crucial for its clinical deployment. In this study, we suggest a machine learning framework for the diagnosis of lung disease that is both interpretable and secured by blockchain using biosensor signals. The proposed framework involves preprocessing the biosensor data, clustering using the K-means algorithm, classification with the XGBoost model, interpretability analysis with SHAP, and data security through blockchain integration. Before classification, clustering was added to minimize intra-class variability, thereby enhancing the separation of patient cluster patterns. The integration of Blockchain with IPFS-based storage was integrated to provide a secure, tamper-proof, and privacy-preserving environment for handling healthcare information. The proposed model yielded 95.80% accuracy, 0.95 F1-score, and 0.98 ROC-AUC when trained on 90% of the data and tested on 10% of the data. Stratified 10-fold cross-validation (10-fold CV) had an accuracy of 94.12% and an ROC-AUC of 0.97. Noise perturbation analysis kept the accuracy high at 92.70% distortion for 10% distortion. The implementation of blockchain resulted in 100% data integrity and low latency (1.8 s). The proposed lung-diagnosis scheme using plasmonic biosensor data is accurate, interpretable, computationally efficient, and secure.