Medical Imaging DICOM Project

Project Overview

This project aims to develop a web-based platform for efficient visualization and analysis of medical imaging (DICOM) data.
To integrate AI diagnostic models from medical AI research into real clinical workflows,
we implemented an interface that allows simultaneous comparison and review of AI diagnostic results and original images.

Key Features

• 🩻 DICOM Viewer
  Supports multi-frame (CT/MRI) images, zoom and pan functions, and window-level adjustments
• 📏 Distance Measurement Tool
  Enables precise millimeter-scale measurement using pixel spacing metadata
• 🤖 AI Prediction Visualization
  Displays predictions from PyTorch-based deep learning models as bounding box overlays
• 🔄 Result Comparison Mode
  Switch between original and AI-analyzed images with a single button
• 🌐 Web-Integrated System
  Built with FastAPI backend and Vue 3 frontend, enabling real-time data communication

Tech Stack

• Frontend: Vue 3, TypeScript, Pinia, SVG-based interactive tools
• Backend: FastAPI, Python, Uvicorn, PM2 (AWS EC2 deployment)
• AI Model: PyTorch, OpenCV, NumPy
• Others: DICOM metadata parsing (x00280030, x00280010, etc.), REST API communication

Project Significance

In medical imaging AI research, it is crucial not only to improve model accuracy
but also to develop tools that medical professionals can actually use.
This project goes beyond experimental AI development —
it is evolving into a “bridge” system connecting medical research and clinical practice.