JSPM

  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 13
  • Score
    100M100P100Q48864F
  • License MIT

Reads any type of medical image in the browser.

Package Exports

  • med-img-reader

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (med-img-reader) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

med-img-reader

Read or write a 3D or 2D image with one or multiple components in a variety of formats. DICOM, NIFTI, NRRD, JPG, PNG, MHD Viewer example

Installation


npm install med-img-reader

Usage example

In NODE environment

Read a single image file


const MedImgReader = require('med-img-reader');

//Image with one or multiple components in any format
var inputImage = '/path/to/input{.nrrd,.nii.gz,.jpg,.png,.dcm}'
var outputImage = '/path/to/output{.nrrd,.nii.gz,.jpg,.png,.dcm}'


const medImgReader = new MedImgReader();
medImgReader.SetFilename(inputImage);
medImgReader.ReadImage();
var image = medImgReader.GetOutput();
console.log("Image:", image);


const medImgWriter = new MedImgReader();

medImgWriter.SetInput(image);
medImgWriter.SetFilename(outputImage);
medImgWriter.WriteImage();

This is an example of an object return by the reader, which is compatible with itk.js


image = {
    imageType: {
      dimension: 2,
      componentType: 'uint16_t',
      pixelType: 1,
      components: 1
    },
    name: 'Image',
    origin: [ 0, 0 ],
    spacing: [ 0.148489, 0.148489 ],
    direction: { rows: 2, columns: 2, data: [ 1, 0, 0, 1 ] },
    size: [ 256, 256 ],
    data: Uint16Array [...]
}

Convert the image to a Tensor from tensorflow tfjs


const tf = require('@tensorflow/tfjs-node');//Or tfjs in browser or tfjs-node-gpu if in linux

tf.tensor(
    Float32Array.from(image.data), 
    [...[...image.size].reverse(), image.imageType.components]
));

Read a DICOM series


var inputDirectory = '/path/to/series/directory'
var outputImage = 'out.nrrd';

const medImgReader = new MedImgReader();
medImgReader.SetDirectory(inputDirectory);
medImgReader.ReadDICOMDirectory();

var image = medImgReader.GetOutput();

const medImgWriter = new MedImgReader();
medImgWriter.SetInput(inputImage);
medImgWriter.SetFilename(outputImage);
medImgWriter.WriteImage();

In browser environment

React component

If you are going to use this in the browser, the build time will be long so be patient. This library is compiled using emscripten and it bundles a file system with the FS library.

Here is an example for a React component:


import React, { Component } from 'react'

const axios = require('axios');
const MedImgReader = require('med-img-reader');

export default class ExampleComponent extends Component {

  constructor(){
    super();

    this.state = {
      itkImage: {}
    }

    const self = this;
    var medImgReader = new MedImgReader();

    var filename = '/brain.png';

    axios({
      method: 'get',
      url: filename,
      responseType: 'blob'
    })
    .then(function(brain){
      var blob = brain.data;
      return blob.arrayBuffer()
      .then((arr)=>{
        medImgReader.WriteFile(filename, arr);//We add the file to the FS filesystem
        medImgReader.SetFilename(filename);//Set the file name 
        medImgReader.ReadImage();
        return medImgReader.GetOutput();
      })
      
    })
    .then((itkImage)=>{
      self.setState({...self.state, itkImage})
    })
  }

  render() {

    const {
      itkImage
    } = this.state;

    var copyImg = {...itkImage, data: []};

    return (
      <div className={styles.test}>
        {JSON.stringify(copyImg)}
      </div>
    )
  }
}

Display the image using vtk.js

Example is here react-med-img-viewer