AI-Driven Early Detection of Breast Cancer

Use Case:

Early detection of breast cancer using sensor signal processing and ML models; non-invasive testing from urine samples collected in hospitals and clinics.

Outcome:

• High-accuracy detection; Developed a hybrid ML model (convolutional + dense + attention layers);
• Implemented ensemble prediction strategy for robustness;
• Deployed model via FastAPI/Docker on AWS cloud infrastructure.

Ecosystem Support:

Full-stack technical mentoring from Deduce Data Solutions via StairwAI; iterative meetings on data science, ML development, infrastructure and knowledge transfer

AI Relevance:

• Real-world health AI deployment with measurable impact;
• AI-powered diagnostic aid aligned with EU priorities on digital health and inclusion;
• Modular architecture supports reproducibility and federated/online learning applications.

Summary:

The Blue Box, a Spanish biomedical startup, embarked on an ambitious mission to make early detection of breast cancer accessible and non-invasive. Their solution uses sensor data from urine samples to detect volatile organic compounds (VOCs) associated with the disease, processed through machine learning algorithms.

With support from the StairwAI programme, and collaboration with Deduce Data Solutions, the team significantly enhanced the robustness of their predictive model. They transitioned from basic dense neural networks to a hybrid architecture combining convolutional and dense layers with attention mechanisms, optimizing for real-world performance. The AI model achieved strong results—87.33% sensitivity and 83.32% accuracy—and was deployed via a Dockerized API on AWS, making it scalable and cloud-ready. By digitizing diagnostic support tools, The Blue Box lowers the barrier for early screening, especially in low-resource or remote healthcare settings. Their model architecture also lays the groundwork for online learning, allowing future adaptive refinement as more data is collected.

This initiative showcases how AI can profoundly impact public health, particularly for women’s health innovation, offering a replicable and socially meaningful use case for the AI-on-Demand ecosystem.