This paper presents a computer-aided diagnosis (CAD) system for detecting COVID-19 using AI-assisted CT imaging analysis. The system, named RACNet, is a deep learning model that integrates classification and segmentation components to analyze 3D CT scans and provide infection probability. The system is designed to be deployed on both cloud and edge environments, ensuring scalability and security. The system uses a microservices architecture to handle sensitive data on the edge and computationally intensive tasks on the cloud. The RACNet model is trained on a large dataset, COV19-CT-DB, containing 7,756 annotated 3D CT scans, with 1,661 cases of COVID-19 and 6,095 non-COVID-19 cases. The model is enhanced with routing and feature alignment steps to dynamically select specific RNN outputs for decision-making. The system also includes a mechanism for extracting latent variables and generating anchor sets to improve explainability and provide confidence levels for diagnoses. The system is designed to be efficient, secure, and scalable, with the ability to handle new datasets and retrain models without affecting existing functionality. The system provides a user-friendly interface for doctors to upload DICOM images and receive diagnoses with explanations. The paper also discusses the ethical compliance of the system and the future work plans for improving the model through user feedback and retraining. The system is expected to reduce the workload of physicians and improve the accuracy of COVID-19 detection.This paper presents a computer-aided diagnosis (CAD) system for detecting COVID-19 using AI-assisted CT imaging analysis. The system, named RACNet, is a deep learning model that integrates classification and segmentation components to analyze 3D CT scans and provide infection probability. The system is designed to be deployed on both cloud and edge environments, ensuring scalability and security. The system uses a microservices architecture to handle sensitive data on the edge and computationally intensive tasks on the cloud. The RACNet model is trained on a large dataset, COV19-CT-DB, containing 7,756 annotated 3D CT scans, with 1,661 cases of COVID-19 and 6,095 non-COVID-19 cases. The model is enhanced with routing and feature alignment steps to dynamically select specific RNN outputs for decision-making. The system also includes a mechanism for extracting latent variables and generating anchor sets to improve explainability and provide confidence levels for diagnoses. The system is designed to be efficient, secure, and scalable, with the ability to handle new datasets and retrain models without affecting existing functionality. The system provides a user-friendly interface for doctors to upload DICOM images and receive diagnoses with explanations. The paper also discusses the ethical compliance of the system and the future work plans for improving the model through user feedback and retraining. The system is expected to reduce the workload of physicians and improve the accuracy of COVID-19 detection.