Design and Analysis of an Integrated AI-Driven andBlockchain-Enabled Connected Framework forNext-Generation Oncology Applications

A wide variety of new cancer diagnostic methods have emerged alongside existing ones, though many still suffer from accuracyand accessibility issues. With the ever-increasing demand for timely and accurate cancer detection, the limitations need to beaddressed to benefit patients. The research proposes a novel framework that leverages artificial intelligence (AI) and integratesit with blockchain technology to enhance oncology applications by addressing critical challenges in cancer diagnosis. As webecome increasingly reliant on AI for medical diagnosis, data privacy, and transparency, a regulatory framework is becomingincreasingly important. We propose a connected architecture based on federated learning that enables collaborative model trainingacross institutions while keeping patient data decentralized and secure. By combining convolutional neural networks (CNNs)with blockchain for provenance tracking, we can achieve diagnostic accuracy, consistency, streamlined workflows, and increasedstakeholder trust. A model optimised across diverse datasets without compromising patient privacy may accelerate the clinicalapplication of AI-based diagnostic tools in oncology. This framework was validated using the PatchCamelyon and BreaKHisdatasets in a non-IID federated setting. The EfficientNet-B0 baseline under centralized training achieved the AUC-ROC scores of0.9557 and 1.0000 in the validation and test datasets, respectively. SimCLR-based fine-tuning yielded AUC-ROC scores of 0.9148and 0.9928, which are highly competitive. In the federated setting, FedProx achieved the best performance (μ = 0.01) with anAUC-ROC of 0.9241 on PCam and an F1-score of 0.9600 on BreaKHis. The integration of blockchain technology successfullymaintained the model’s utility with negligible loss. Thus, a federated oncology framework that can be secure, privacy-preserving,and auditable is certainly possible.