NeuroScan

Inspiration

Our Inspiration came from hearing the stories of Medicine Students and hearing about the struggles specifically in the Radiology department. When we were told about these struggles we felt that we could implement an idea which could cut down massively on these delays by implementing AI.

What it does

This AI takes in a CT brain scan, identifies it and classifies it into one of the 4 main types of haemorrhages; Extradural haematoma, Subdural haematoma, Subarachnoid haemorrhage and Intracerebral haemorrhage. We designed the website with Radiologists in mind, keeping a minimilistic style to avoid clutter and confusion in an emergency.

How we built it

As a result of the time constraints, we decided to build this idea using Claude code for the backend and the Lovable website design for the UI/ frontend.

Challenges we ran into

We found that training the AI to respond accurately to the different types of haemorrhaging. In the beginning the AI could only correctly identify

What we learned

Building NeuroScan taught us the importance of precision when lives are on the line. We deepened our understanding of both radiology and machine learning. We also learned how to work efficiently under time pressure and how to bridge the gap between clinical knowledge and software development.

What's next for NeuroScan

We want to expand NeuroScan's capabilities beyond classification, adding severity scoring, haematoma volume estimation, and priority flagging to help radiologists triage the most critical cases first. Longer term, we aim to validate the model against larger clinical datasets, pursue partnerships with hospitals and attempt to sell them this idea on a per month/ per patient rate.

Built With

• claude
• openai
• python
• typescript