Clariscan

• 

  HomePage

• 

  Simplified Report generated

Note : Included demo video and audio regarding codex and chatgpt in the youtube video.

Inspiration

ClariScan was inspired by a common problem many patients face after receiving their radiology reports. The few hours or days between receiving a report and consulting a doctor are often filled with uncertainty, as patients struggle to understand complex medical terminology and turn to the internet for answers, leading to confusion, misinformation, and unnecessary anxiety. We wanted to bridge this gap by creating a solution that empowers patients with clear, reliable, and easy-to-understand explanations during this crucial waiting period without replacing professional medical advice.

What it does

ClariScan is a patient-friendly web application that converts complex radiology reports into structured, easy-to-understand summaries. Users can upload a PDF or image of their report, and ClariScan generates a simplified explanation that includes why the scan was performed, the key findings in plain language, an overall summary, a risk indicator, and a safety disclaimer reminding users that the report is not a medical diagnosis. The generated summary can also be downloaded or printed for future reference during the doctor's consultation.

How we built it

We built ClariScan as a complete end-to-end solution using OCR to extract text from uploaded PDFs and images, followed by locally hosted Large Language Models (LLMs) through Ollama to generate accurate and patient-friendly explanations. The application was designed with a simple and intuitive interface, ensuring that users of all technical backgrounds can easily upload reports and receive understandable summaries while keeping their medical data private.

Challenges we ran into

One of our biggest challenges was ensuring that AI-generated explanations remained accurate, reassuring, and easy to understand without crossing the boundary into medical diagnosis. We also worked on creating meaningful risk indicators while avoiding unnecessary alarm, handling different report formats during text extraction, and maintaining a smooth user experience with locally hosted AI models.

Accomplishments that we're proud of

We're proud of building a fully functional MVP that addresses a real healthcare problem often overlooked in the patient journey. By focusing on the time between receiving a radiology report and meeting a doctor, ClariScan helps patients feel more informed and less anxious. We're also proud of using self-hosted LLMs, enabling privacy-focused AI without relying on external APIs or incurring inference costs.

What we learned

This project taught us how AI can simplify complex medical information while emphasizing the importance of responsible AI in healthcare. We gained hands-on experience integrating OCR, locally hosted LLMs, and web technologies into a seamless workflow. Most importantly, we learned that good AI solutions are not just technically accurate, they must also build trust, improve accessibility, and provide value without replacing professional expertise. We specifically used codes to build our basic application and further improvised it with the help of Chatgpt 5.6 . By using codex and chatgpt , we were able to accelerate our work and create a fully working project within the given period of time.

What's next for ClariScan

Our roadmap includes expanding support beyond radiology reports to laboratory reports and other medical documents, introducing multilingual summaries for greater accessibility, and adding secure cloud-based report history for premium users. We also plan to integrate with hospitals and diagnostic centers through APIs, enabling ClariScan to become a trusted companion that helps patients better understand their medical reports before every consultation.

Built With

• angular.js
• ocr
• ollama-llm
• pytesseract
• python