Scholarlink

<p>Multi-organ rare disease prediction uses deep learning models to analyze medical scans of organs like the heart, liver, kidney, and lungs, enabling early detection and accurate diagnosis of rare diseases. This study presents a novel approach for predicting multi-organ rare diseases, including heart, liver, kidney, and lung cancer, by integrating federated learning (FL) with natural language processing (NLP) in a cloud-based environment. Rare disease prediction faces significant challenges due to data scarcity, privacy concerns, and limited generalization across multiple organs. To address these issues, the proposed framework employs advanced preprocessing techniques such as missing data handling, normalization, and SMOTE to ensure high-quality, consistent input. Feature extraction is enhanced using Word2Vec and BERT for data, while an improved graph neural network (I-GNN) is employed to capture complex spatial relationships within medical data. For disease classification, an ensemble model combining convolutional neural networks (CNN), long short-term memory networks (LSTM), and Random Forest with hard voting is utilized. This integration leverages the strengths of various models to improve prediction accuracy and robustness across diverse multi-organ rare disease diagnoses. The cloud-based framework facilitates secure and scalable processing of sensitive medical data, ensuring efficient collaboration and privacy preservation. The proposed model is implemented using the Python NumPy library, and its performance is evaluated for both 70% and 80% of training data for the metrics Accuracy of 0.9838, 0.9932, Precision of 0.9747, 0.9907, and also the Sensitivity, F1-Score, G-Mean, MSE, MAE, BLEU Score, and MCC have consistently achieved higher accuracy than existing methods.</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

NLP-driven federated learning framework for multi-organ rare disease prediction

  • 期刊论文

  • 发表日期：

    2026年1月12日

查看全文
  • Shahnazeer Chembalakkat Kunjumohamad,
  • Sureshkumar Govindaraj

摘要

Multi-organ rare disease prediction uses deep learning models to analyze medical scans of organs like the heart, liver, kidney, and lungs, enabling early detection and accurate diagnosis of rare diseases. This study presents a novel approach for predicting multi-organ rare diseases, including heart, liver, kidney, and lung cancer, by integrating federated learning (FL) with natural language processing (NLP) in a cloud-based environment. Rare disease prediction faces significant challenges due to data scarcity, privacy concerns, and limited generalization across multiple organs. To address these issues, the proposed framework employs advanced preprocessing techniques such as missing data handling, normalization, and SMOTE to ensure high-quality, consistent input. Feature extraction is enhanced using Word2Vec and BERT for data, while an improved graph neural network (I-GNN) is employed to capture complex spatial relationships within medical data. For disease classification, an ensemble model combining convolutional neural networks (CNN), long short-term memory networks (LSTM), and Random Forest with hard voting is utilized. This integration leverages the strengths of various models to improve prediction accuracy and robustness across diverse multi-organ rare disease diagnoses. The cloud-based framework facilitates secure and scalable processing of sensitive medical data, ensuring efficient collaboration and privacy preservation. The proposed model is implemented using the Python NumPy library, and its performance is evaluated for both 70% and 80% of training data for the metrics Accuracy of 0.9838, 0.9932, Precision of 0.9747, 0.9907, and also the Sensitivity, F1-Score, G-Mean, MSE, MAE, BLEU Score, and MCC have consistently achieved higher accuracy than existing methods.