This repository contains a web-based platform for processing DICOM (Digital Imaging and Communications in Medicine) images. Users can upload DICOM images, and the system extracts, normalizes, and thresholds the pixel data. After processing, the system displays the volume (in mm³) of the thresholded pixels.

• Medical Imaging Coding Assessment
  • Overview
  • Table of Contents
    • Web Interface
    • Backend
    • Docker
    • Code Quality and Testing
  • Setup
    • Prerequisites
    • Instructions
    • Future Work
    • Not Implemented But Thought Of

• The web interface is implemented using Vue.js and provides a simple page with a drop zone for uploading DICOM images.
• Upon the click on the dropzone it also provides uploadfile menu.
• After processing, the number of pixels above the threshold is displayed.

• The backend is built with FastAPI, providing REST endpoints for communication with the web interface.
• The backend uses the Python package pydicom to extract pixel data from DICOM files.
• The pixel data is normalized to a range of [0, 1].
• Thresholding is applied to the pixel data using a threshold value (defaulting to 0.5 if not provided).
• The volume of thresholded pixels in mm³ is returned.

• Docker images are provided for both the backend and frontend.
• A docker-compose.yml file is included for running the solution in a Docker environment.

• The code is well-documented (by choosing variables names carefully) and partially tested (With additional suggestions).
• Consideration has been given to various edge cases that might occur when processing DICOM files.
• Not Implemented But Thought Of

• Docker and Docker Compose must be installed on your system.

1. Clone this repository to your local machine.

   git clone [email protected]:xbankov/dicom_webapp.git
   
   cd dicom_webapp
   
   docker-compose up --build

• implement dynamic thresholding techniques to improve accuracy for different types of DICOM images
• advanced dicom processing using neural netword (nnunet, other semantic segmantation networks, image registration, etc ...)
• user-friendly DICOM viewer for enhanced user interaction ( borrowing from 3Dslicer)
• Implementing anomaly detection, is another potential future improvement.
• Removing CT table (part of my master thesis at DKFZ)
• STYLES - a small touch-up would be beneficial, mostly for the easy of use
• Use PHYSICAL SPACING-aware medical processing library - like SITK

• tests for validity of dicom metadata AND data, like intensities in pixel_array check or spacing check.
• tests for the not .dcm files, e.g. (.txt)
• tests for the actual content of the CT scan