1,2Department of Computer Science, University Malaysia of Computer Science & Engineering (UNIMY)
3Kyber Medical College, Peshawar, Pakistan
4Istanbul Atlas University, Hamidiye, Turkey
5Department of Neurosurgery, University of Colorado | Denver Health Medical Center/Children's Hospital, Colorado, USA
6Department of Neurosurgery, Ziauddin University, Karachi, Pakistan
Email: Soobiasaeed1@gmail.com, habibollah@unimy.edu.my, izazriaz243@gmail.com, halahshah01@gmail.com, Muhammad.Riaz2@dhha.org, Mohsin.qadeer@gmail.com
*Correspondence Author: Soobia Saeed, Email: soobiasaeed1@gmail.com
The healthcare sector has undergone an accelerated digital transformation which has resulted in the creation of large volumes of sensitive medical information and has simultaneously made it possible to use the power of AI to predict diseases and have serious data privacy, security, and compliance issues arise. Centralized healthcare systems, which are the norm, are finding it harder to secure their data and are against the possible legal and ethical battles posed by the strict regulations such as HIPAA and GDPR. This study presents an innovative solution of an AI-based disease prediction system that is meticulously designed to work with federated learning and a decentralized healthcare management framework. The main feature of the solution is that it allows different hospital networks to jointly train the same model without the need of sending the actual patient data among the hospitals thus, it is not only able to maintain the privacy of the data but also the power over its usage. The system combines federated learning with machine learning and deep learning-based models, which consist of, among others, Convolutional Neural Networks (CNNs) for medical imaging analysis, Support Vector Machines (SVMs) for structured clinical data, and Random Forest classifiers for multimodal prediction. Secure aggregation, encryption methods, and role-based access management are implemented to protect data authenticity and privacy during the course of the system operation. The application of such data sets that are freely accessible and stripped of personal identifiers is meant to create a representation of the real-world multi-institutional healthcare settings and this is also a way to demonstrate the proposed method's feasibility, scalability, and robustness. Decentralized federated learning models, according to the experiment outcomes, are capable of performing at the same level as centralized methods with the exception that they come with much lesser privacy risk. The authors of the paper emphasize the potential that the use of federated learning–based decentralized healthcare systems can have in the areas of early disease detection support, clinical decision making, and the deployment of responsible AI that is compatible with the ethical standards prevailing in the modern healthcare ecosystem.
Keywords: Federated Learning with decentralized healthcare, disease prediction, and privacy-preserving AI.
How to cite this article: Saeed S, Haron H, Riaz I, Shah HK, Riaz M, Qadeer M. A Federated Learning-based project for predicting disease diagnoses using Artificial Intelligence. Int J Drug Deliv Technol. 2026;16(10s): 975-984. DOI: 10.25258/ijddt.16.10s.113
Source of support: Nil.
Conflict of interest: None