EDGE-TO-CLOUD AI MODEL
DEPLOYMENT FOR LUNG CANCER
PREDICTION SYSTEM
INTRODUCTION
• The Edge-to-Cloud AI Model Deployment Framework is designed to enhance real-
time AI inferencing by leveraging edge computing while maintaining seamless cloud
integration.
• This framework enables AI models to be deployed on edge devices, allowing for
low-latency processing and efficient decision-making at the edge.A monitoring and
feedback loop is integrated to track model performance and refine AI models based
on real-world data.
• This framework offers a scalable, secure, and efficient AI deployment system
applicable across multiple industries.
2
PROBLEM STATEMENT
• Traditional cloud-only AI systems suffer from latency
• Hospitals in rural areas lack GPU infrastructure
• Medical AI models require frequent updates
• Need for real-time, reliable & scalable framework
• Need for DevOps automation for model lifecycle
3
PROJECT MOTIVATION
• Use edge devices for low-latency inference
• Use cloud Graphic Processing Units for large-scale training
• Automate deployments using DevOps pipelines
• Enhance reliability with OTA model updates
• Improve medical outcomes with early prediction
4
OBJECTIVES
• Develop a real-time lung cancer prediction system
• Deploy optimized AI model on edge devices
• Build cloud-based training & monitoring pipeline
• Enable OTA model updates across all edge devices
• Implement a DevOps-driven CI/CD workflow
• Ensure scalability, reliability, and security
5
LITERATURE REVIEW
•Cloud–Edge Collaboration for Medical AI: Research shows that combining cloud
computing for heavy model training and edge devices for real-time inference improves
speed, privacy, and reliability in cancer prediction systems.
•Adaptive Deployment Improves Performance: Studies propose algorithms that
dynamically place services across cloud and edge based on speed, latency, and
resource conditions, significantly reducing delay and balancing workload.
•Distributed ML Enhances Efficiency: Frameworks like InTec demonstrate that
distributing machine learning tasks across Things–Edge–Cloud layers reduces latency,
network usage, and energy consumption for real-time applications.
6
SYSTEM ARCHITECTURE
• CT scan image is sent to the edge device, which receives and processes
medical images instantly.
• The edge device runs the AI inference model and gives real-time lung cancer
predictions without needing cloud processing.
• The IoT layer connects the edge to the cloud, handling logging,
synchronization, and OTA model updates.
• The cloud layer trains the model, stores datasets, runs CI/CD pipelines, and
monitors system performance.
• The user interface displays prediction results and monitoring information,
making the system easy for doctors to use.
7
SYSTEM ARCHITECTURE
8
EDGE LAYER DESIGN
• Purpose:
• Perform low-latency predictions on-device
• Functions:
• Image preprocessing
• AI inference (INT8 / FP16 model)
• Offline operation
• Sends logs to cloud
• Devices:
Jetson Nano, Raspberry Pi
9
CLOUD LAYER DESIGN
• Purpose:
• Heavy processing & automation
• Functions:
• Model training on cloud
• Dataset storage
• DevOps CI/CD pipelines
• Monitoring & analytics
• Tools:
10 AWS, Azure, Docker, Kubernetes
IOT SYNCHRONIZATION LAYER
• Purpose:
Handle communication between edge & cloud
• Functions:
• Secure MQTT/HTTPS messages
• OTA model updates
• Device configuration (Digital Twin)
• Logging and telemetry
• Platforms:
AWS IoT Core
11
DEVOPS WORKFLOW
Pipeline Steps:
• Code commit triggers CI/CD
• Model testing & accuracy validation
• Docker image build
• Push to registry
• OTA deployment to edge devices
• Monitoring & feedback loop
Tools:
GitHub Actions, Jenkins, Kubernetes, Terraform, Grafana
12
TOOLS AND TECHNOLOGIES
Infrastructure Automation
CI/CD Tools
• Terraform
• GitHub Actions
• Ansible
• Jenkins
Monitoring & Logging
Containerization
• Prometheus
• Docker
• Grafana
• Docker Compose
Cloud DevOps
Orchestration
• AWS CodePipeline
• Kubernetes
13 • Azure DevOps
IMPLEMENTATION DEATILS
•Used LIDC–IDRI (Lung Image Database Consortium–Image Database
Resource Initiative) dataset with CT image preprocessing.
•Trained CNN (Convolutional Neural Network) model using PyTorch and
optimized it with ONNX (Open Neural Network Exchange) and TensorRT (Tensor
Runtime).
•Deployed optimized model on edge devices like NVIDIA Jetson Nano and
Raspberry Pi for real-time inference.
•Implemented CI/CD (Continuous Integration / Continuous Deployment) using
GitHub Actions and OTA (Over-The-Air) updates for model delivery.
14
ADVANTAGES
• Real-time prediction
• Low latency inference
• Continuous model improvement
• Fully automated DevOps pipeline
• Scalable across multiple hospitals
• High reliability and security
15
RESULT AND ANALYSIS
•The model correctly identified lung cancer with high accuracy.
• Edge device gave fast, real-time predictions with low delay.
• DevOps pipeline enabled quick and automatic model updates.
• Overall system became efficient, reliable, and ready for real use.
16
CONCLUSION AND FUTURE SCOPE
• Successfully built edge-to-cloud AI framework
• Achieved real-time lung cancer detection
• DevOps ensured automated, reliable deployment
• Future scope: 3D volumetric analysis, federated learning, HIS integration
17
REFERENCES
• [1] A Cloud-Edge Collaboration Framework for Cancer Survival Prediction to
Develop Medical Consumer Electronic Devices, IEEE Xplore, 2024. Available:
[Link]
• [2] Z. Hu, S. Chen, H. Rao, C. Hong, O. Huang, X. Xu, and G. Jia, “An Adaptive Service
Deployment Algorithm for Cloud-Edge Collaborative System Based on Speedup Weights,”
The Journal of Supercomputing, vol. 16, pp. 23177–23204, 2024. doi: 10.1007/s11227-024-
06339-8.
• [3] H. Larian and F. Safi-Esfahani, “InTec: Integrated Things-Edge Computing—A
Framework for Distributing Machine Learning Pipelines in Edge AI Systems,” Computing,
vol. 107, no. 1, pp. 1–19, 2025. doi: 10.1007/s00607-024-01388-6.
• [4] Deployment of Artificial Intelligence Models on Edge Devices: A Tutorial Brief,
IEEE Internet of Things Journal, 2023. Available:
[Link]
• [5] Edge AI as a Service: Configurable Model Deployment and Delay-Energy
Optimization With Result Quality Constraints, IEEE Transactions on Cloud Computing,
2023. Available: [Link]
18
THANK YOU