Federated IoMT-assisted ensemble learning for automated pulmonary CT analysis

Abstract Pulmonary diseases such as COVID-19, pneumonia, and tuberculosis continue to impose significant challenges on global healthcare systems, emphasizing the need for rapid, reliable, and privacy-aware diagnostic frameworks. Chest computed tomography (CT) imaging has demonstrated remarkable effectiveness in the early screening and assessment of pulmonary abnormalities. In recent years, the integration of deep learning and Internet of Medical Things (IoMT)-enabled healthcare technologies has emerged as a promising solution for automated medical image analysis and intelligent remote healthcare services. This study proposes a Federated IoMT-Assisted Ensemble Deep Learning framework for automated pulmonary CT analysis. The framework integrates an IoMT-enabled healthcare deployment architecture, federated learning mechanisms, and heterogeneous ensemble transfer learning models to facilitate secure, scalable, and privacy-preserving pulmonary disease diagnosis. A Federated Averaging (FedAvg)-based distributed learning strategy is implemented to support privacy-preserving collaborative model training across multiple healthcare institutions without directly sharing sensitive patient data. The proposed ensemble architecture utilizes three state-of-the-art pre-trained convolutional neural network (CNN) models, namely ResNet152V2, DenseNet201, and InceptionResNetV2, for robust feature extraction and multi-class pulmonary disease classification. Transfer learning, adam optimization, dropout regularization, and early stopping mechanisms are employed to improve model convergence, reduce overfitting, and enhance generalization capability. The framework is evaluated using a multi-class chest CT dataset consisting of COVID-19, pneumonia, tuberculosis, and normal cases collected from multiple publicly available datasets. Extensive experimental analysis is conducted using several performance metrics, including accuracy, precision, recall, F1-score, and Area Under the Curve (AUC). Experimental results demonstrate that the proposed federated ensemble framework consistently outperforms several state-of-the-art pulmonary disease classification approaches. The proposed model achieves a classification accuracy of 98.96%, with precision, recall, F1-score, and AUC values of 98.55%, 98.57%, 98.56%, and 98.56%, respectively. Furthermore, the federated learning-assisted framework demonstrates the feasibility of privacy-preserving distributed pulmonary disease diagnosis while supporting secure healthcare communication and collaborative intelligence across IoMT-enabled healthcare ecosystems. The obtained results indicate that the proposed framework can serve as an efficient, scalable, and reliable decision-support system for next-generation intelligent pulmonary disease screening and remote healthcare applications. The proposed framework contributes to Sustainable Development Goal (SDG) 3 (Good Health and Well-Being) by supporting early and accurate pulmonary disease diagnosis, improving access to intelligent healthcare services, and facilitating privacy-preserving remote healthcare delivery. Furthermore, the integration of federated learning and IoMT-enabled infrastructure aligns with SDG 9 (Industry, Innovation and Infrastructure) by promoting secure, scalable, and innovative digital healthcare solutions.