Dark Mode Landing Page
Dark Mode Photo Upload
Dark Mode Image Result
Dark Mode Image History
Dark Mode About
Light Mode Landing Page
Light Image Upload
Light Mode Image Result
Light Mode Image History
Light Mode About
Cloud Pipeline
Confusion Matrix
Results

Derma – AI-Assisted Skin Cancer Screening

Inspiration

Skin cancer is highly treatable when detected early, yet dermatologists face increasing patient demand and limited time per evaluation. We were inspired by the idea of transforming a smartphone into a clinical decision-support tool that delivers an instant AI-powered secondary opinion during examinations.

More broadly, Derma reflects the growing role of artificial intelligence in the ever-changing healthcare environment. As clinical workloads increase and technology advances, AI has the potential to support faster decision making, improve consistency, and expand access to high-quality care.

Derma is built to strengthen clinical confidence and improve efficiency, not replace medical expertise.

What It Does

Derma is a mobile AI application designed for dermatologists.

It allows clinicians to:

Capture or upload dermatoscopic images from their phone
Receive a classification of potential cancer or non-cancer
View a confidence score
Store and review structured analysis results

The app functions as a real-time AI assistant within a dermatologist’s workflow.

How We Built It

Model

We used YOLO26s-cls, a lightweight image classification model optimized for fast and efficient inference.

Trained on approximately 10,000 dermatoscopic images
Training performed on the URI Unity Cluster
Tuned for deployment in a cloud-based production environment

Images are preprocessed using OpenCV before being passed into the trained model for classification.

Backend and Cloud Infrastructure

We designed a scalable and production-ready cloud architecture:

Flutter (Mobile) for cross-platform iOS and Android development
Flask (Python) backend to handle network API requests and log debugging information
Docker to containerize the backend and model environment
Artifact Registry to store and manage container images
Google Cloud Run to deploy and auto-scale containers
Firebase Storage to hold uploaded dermatoscopic images and model training weights
Firestore (NoSQL database) to store structured results data
Firebase Authentication for secure access control

System Flow

Images captured in the mobile app are uploaded to Firebase Storage.
The Flask API receives a request and retrieves the image.
The YOLO26s-cls model performs classification.
Results are written to Firestore.
The mobile app retrieves and displays the results.

Docker ensures consistent environments across development and production. Containers are pushed to Artifact Registry, and Cloud Run automatically scales them based on incoming traffic.

Challenges We Faced

Training Stability

Medical image data required careful preprocessing and augmentation to prevent overfitting.

Cloud Deployment Complexity

Managing Docker builds, Artifact Registry images, and Cloud Run deployments required careful configuration and testing.

Latency Optimization

Inference had to be fast enough to support real-time clinical use.

Clinical Positioning

Ensuring the system is clearly framed as decision support rather than diagnosis required thoughtful design and communication.

What We Learned

Production AI systems require strong infrastructure and DevOps knowledge in addition to model development.
Lightweight architectures like YOLO26s-cls can balance speed and performance effectively.
Logging and observability through the Flask API layer are essential for reliability and debugging.
AI has significant potential in healthcare, including diagnostic support, workflow automation, remote screening, and decision augmentation.

Healthcare is rapidly evolving, and AI systems like Derma demonstrate how intelligent tools can integrate into clinical practice to improve efficiency and support better outcomes.

What’s Next

Our long-term goal is to partner with companies that manufacture dermatoscopes or smartphone dermatoscope attachments.

By combining:

High-quality dermatoscopic imaging hardware
AI-powered classification software
Real-time mobile deployment

Derma could evolve into a professional clinical support platform. Dermatologists would capture images using dermatoscope attachments and instantly receive AI-assisted analysis directly on their phone as a seamless secondary opinion tool.

Our vision is to build scalable and trustworthy AI that enhances diagnostic confidence and contributes to the future of intelligent healthcare systems.