技术笔记：一个用于影像组学综合的MATLAB平台

The Computational Environment for Radiological Research (CERR)

作者来自纽约大学、剑桥大学，发表在medical physics

目的：医学图像量化分析，包含数据导入、transformation，management，feature calculation, image segmentation, and visualization. 所有的 radiomics 计算被向量化获得加速。此外a test suite is provided for reconstruction and comparison with radiomics features computed using other software platforms such as the Insight Toolkit (ITK) and PyRadiomics。

一些用于影像组学的开源软件

• ITK
• PyRadiomics, extract radiomics with a plugin for 3D Slicer, 缺乏the calculation of radiomics maps and DICOM-RT input of anatomical structures.
  这些都是radiomics library而不是integrated platform

CERR旨在克服上述工具的缺陷，它提供了tests for its radiomics features against other open-source software，同时也compute higher-order texture features using vectorized implementations，产生了显著的加速。用纯matlab开发，making it agnostic to the operating system and MATLAB versions.

1. Architecture

CERR关键的六大功能

• import imaging data with standard formats using different modalities，支持RTOG, DICOM, MHA, NRRD, NIfTI and XML
• delineate and import segmentations for radiomics calculation
  轮廓工具包含铅笔，刷子，橡皮和active contour-based refinement models 以及 Boolean arithmetic to derive new structures from existing ones（比如取并集/交集）。The segment labeler tool可以很方便的让用户graphically score 自动分割结果的不同部分。
• define important parameters for radiomics calculation
• visualize and compare the resulting radiomics maps
• derive and store radiomics values along with imaging data
  把原图跟基于原图的分割结果、影像学特征存储在一起，simplify the bookkeeping. CERR’s radiomics can be called from other applications by just passing the matrices for the image and the labels.
• export the resulting radiomics scalars or maps to any other analysis software

2. Radiomics maps and pre-processing filters

• 快速计算Haralick, Law's and first-order statistical radiomics
• 预处理滤波器：wavelet, sobel, gabor and Laplacian of Gaussian
• 图像插值到用户指定的分辨率
• MR偏移场矫正 （李纯明老师MICO）
  Moreover, CERR’s data structure provides convenient access to images and associated metadata, making it straightforward for users to define custom normalizations or use filters from libraries such as ITK and MATLAB image processing toolbox。

3. Scalar radiomics (6个)

(i) First-order/histogram statistics, (ii) Intensity-volume histogram, (iii) Peak/Valley，(iv) Shape, (v) Size, and (vi) Texture。
位大形变
Scalar texture

• 灰度共生矩阵；
• Neighborhood grey tone difference
• Neighborhood grey level dependence
• Run length
• Size zone
  可以2D/3D计算，特征可以被存储在matlab的数据结构或者输出到一个CSV文件中。

D. Speed-up using matrix algebra