lungct / Git / Diff of /README.md

Models:

SCallahan/

lungct

Downloads: 2

Diff of /README.md [3b2327] .. [e54d1d]

Switch to unified view


---
output: github_document
---

<!-- README.md is generated from README.Rmd. Please edit that file -->

```{r setup, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)
```

## Highlights

* We develop a free and simple segmentation algorithm that is comparable to the proprietary VIDA Diagnostics software

* We create the first publicly available standard lung template using healthy adults, which is available for download via lungct

* We show that the standard lung template allows for improved population-level inference of lung CTs using radiomics

* lungct provides a fully-automated and open-source image processing pipeline for lung CTs, which is accessible to statisticians

## Installation




You can install lungct from github with:

``` r
# install.packages("devtools")
devtools::install_github("muschellij2/lungct")
```


## Segmentation

To segment the lungs from the CT scan: 

``` r
library(lungct)
filename <- "example.nii.gz"
mask <- segment_lung(filename)
```

To segment the left and right lungs from the CT scan: 

``` r
img <- ANTsRCore::antsImageRead(filename)
mask2 <- segment_lung2(img)
```

## Standard Lung


The standard lung was created from N=62 healthy controls from [COPDGene](http://www.copdgene.org/) (50\% female, 95\% white, mean age = 62 years, mean BMI = 28.5)

To load the standard lung: 

``` r
# Read in standard lung template
filepath <- system.file(
    "extdata", "lung_template_mask.nii.gz", 
    package = "lungct")
template <- ANTsRCore::antsImageRead(filepath)
```

## Registration


To register the mask to the standard lung: 

``` r
# Register mask to standard lung
reg <- register_lung_mask(
    moving_mask = mask, 
    fixed_mask = template, 
    moving = img, 
    sides = c("right","left"),
    typeofTransform = "SyN")
```

We recommend the following for lung registration: 

* Separately register the right and left lungs to account for differences in lung shape and size

* Perform registration on lung masks to preserve biological variability in Hounsfield units (HU)

* Use [Symmetric Normalization (SyN)](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2276735/) for nonlinear registration due to its flexibility and success in the [EMPIRE10](https://empire10.grand-challenge.org/) challenge


## Template Creation

For template creation, we follow the iterative method from [Avants](https://www.ncbi.nlm.nih.gov/pubmed/19818860), using iterations of register_lung_mask, get_template, and calculate_DSC from lungct. However, we define convergence as having a Dice similarity coefficient (DSC) between successive iterations of at least 0.99. We recommend using parallelization for the registration step. 

![Template Creation](inst/extdata/templatecreation.png)

```r
# Obtain new template
template_new <- get_template(
    folder_warp = warped_masks, 
    folder_comp = transformations)

# Check convergence     
dice <- calculate_DSC(template_init, template_new)
```

## Radiomics


Radiomics, a field of quantitative imaging where large amounts of features are extracted from medical images, is common for lung CTs. 

Options in lungct: 

* radiomics_slice: Calculation on 2D slices in axial, coronal, or sagittal planes

* radiomics_lung: Calculation on 3D right and left lungs

* RIA_lung: Calculation of advanced radiomic features, such as GLCM and GLRLM [More info](https://github.com/cran/RIA)

```r
# Calculate radiomics
rad <- RIA_lung(
    img, mask, 
    sides = c("right", "left"), 
    features = c("fo", "glcm"), 
    bins_in = 16, equal_prob = FALSE, distance = 1, 
    statistic = "mean(X, na.rm = TRUE)")
```


## COPDGene data usage

For the lung CT template, the [COPDGene data](http://www.copdgene.org/) was used:

"This research used data generated by the COPDGene study, which was supported by NIH grants U01HL089856 and U01HL089897. The COPDGene project is also supported by the COPD Foundation through contributions made by an Industry Advisory Board comprised of Pfizer, AstraZeneca, Boehringer Ingelheim, Novartis, and Sunovion."


	a/README.md		b/README.md
1		1	---
		2	output: github_document
		3	---
		4
2	<!-- README.md is generated from README.Rmd. Please edit that file -->	5	<!-- README.md is generated from README.Rmd. Please edit that file -->
3	lungct	6
4	======	7	```{r setup, include = FALSE}
5		8	knitr::opts_chunk$set(
6	[![Travis-CI Build Status](https://travis-ci.org/muschellij2/lungct.svg?branch=master)](https://travis-ci.org/muschellij2/lungct)	9	collapse = TRUE,
7		10	comment = "#>",
8	The lungct R package develops an image processing pipeline for computed tomography (CT) scans of the lungs.	11	fig.path = "man/figures/README-",
9		12	out.width = "100%"
10	![lungct logo](inst/extdata/logo.png)	13	)
11		14	```
12	![Image Processing Pipeline](inst/extdata/pipeline.png)	15
13		16	## Highlights
14	Highlights	17
15	----------	18	* We develop a free and simple segmentation algorithm that is comparable to the proprietary VIDA Diagnostics software
16		19
17	- We develop a free and simple segmentation algorithm that is comparable to the proprietary VIDA Diagnostics software	20	* We create the first publicly available standard lung template using healthy adults, which is available for download via lungct
18		21
19	- We create the first publicly available standard lung template using healthy adults, which is available for download via lungct	22	* We show that the standard lung template allows for improved population-level inference of lung CTs using radiomics
20		23
21	- We show that the standard lung template allows for improved population-level inference of lung CTs using radiomics	24	* lungct provides a fully-automated and open-source image processing pipeline for lung CTs, which is accessible to statisticians
22		25
23	- lungct provides a fully-automated and open-source image processing pipeline for lung CTs, which is accessible to statisticians	26	## Installation
24		27
25	Installation
26	------------
27
28	You can install lungct from github with:	28	You can install lungct from github with:
29		29
30	``` r	30	``` r
31	# install.packages("devtools")	31	# install.packages("devtools")
32	devtools::install_github("muschellij2/lungct")	32	devtools::install_github("muschellij2/lungct")
33	```	33	```
34		34
35	Segmentation	35
36	------------	36	## Segmentation
37		37
38	To segment the lungs from the CT scan:	38	To segment the lungs from the CT scan:
39		39
40	``` r	40	``` r
41	library(lungct)	41	library(lungct)
42	filename <- "example.nii.gz"	42	filename <- "example.nii.gz"
43	mask <- segment_lung(filename)	43	mask <- segment_lung(filename)
44	```	44	```
45		45
46	To segment the left and right lungs from the CT scan:	46	To segment the left and right lungs from the CT scan:
47		47
48	``` r	48	``` r
49	img <- ANTsRCore::antsImageRead(filename)	49	img <- ANTsRCore::antsImageRead(filename)
50	mask2 <- segment_lung2(img)	50	mask2 <- segment_lung2(img)
51	```	51	```
52		52
53	Standard Lung	53	## Standard Lung
54	-------------	54
55
56	The standard lung was created from N=62 healthy controls from [COPDGene](http://www.copdgene.org/) (50% female, 95% white, mean age = 62 years, mean BMI = 28.5)	55	The standard lung was created from N=62 healthy controls from [COPDGene](http://www.copdgene.org/) (50\% female, 95\% white, mean age = 62 years, mean BMI = 28.5)
57		56
58	To load the standard lung:	57	To load the standard lung:
59		58
60	``` r	59	``` r
61	# Read in standard lung template	60	# Read in standard lung template
62	filepath <- system.file(	61	filepath <- system.file(
63	"extdata", "lung_template_mask.nii.gz",	62	"extdata", "lung_template_mask.nii.gz",
64	package = "lungct")	63	package = "lungct")
65	template <- ANTsRCore::antsImageRead(filepath)	64	template <- ANTsRCore::antsImageRead(filepath)
66	```	65	```
67		66
68	Registration	67	## Registration
69	------------	68
70
71	To register the mask to the standard lung:	69	To register the mask to the standard lung:
72		70
73	``` r	71	``` r
74	# Register mask to standard lung	72	# Register mask to standard lung
75	reg <- register_lung_mask(	73	reg <- register_lung_mask(
76	moving_mask = mask,	74	moving_mask = mask,
77	fixed_mask = template,	75	fixed_mask = template,
78	moving = img,	76	moving = img,
79	sides = c("right","left"),	77	sides = c("right","left"),
80	typeofTransform = "SyN")	78	typeofTransform = "SyN")
81	```	79	```
82		80
83	We recommend the following for lung registration:	81	We recommend the following for lung registration:
84		82
85	- Separately register the right and left lungs to account for differences in lung shape and size	83	* Separately register the right and left lungs to account for differences in lung shape and size
86		84
87	- Perform registration on lung masks to preserve biological variability in Hounsfield units (HU)	85	* Perform registration on lung masks to preserve biological variability in Hounsfield units (HU)
88		86
89	- Use [Symmetric Normalization (SyN)](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2276735/) for nonlinear registration due to its flexibility and success in the [EMPIRE10](https://empire10.grand-challenge.org/) challenge	87	* Use [Symmetric Normalization (SyN)](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2276735/) for nonlinear registration due to its flexibility and success in the [EMPIRE10](https://empire10.grand-challenge.org/) challenge
90		88
91	Template Creation	89
92	-----------------	90	## Template Creation
93		91
94	For template creation, we follow the iterative method from [Avants](https://www.ncbi.nlm.nih.gov/pubmed/19818860), using iterations of register\_lung\_mask, get\_template, and calculate\_DSC from lungct. However, we define convergence as having a Dice similarity coefficient (DSC) between successive iterations of at least 0.99. We recommend using parallelization for the registration step.	92	For template creation, we follow the iterative method from [Avants](https://www.ncbi.nlm.nih.gov/pubmed/19818860), using iterations of register_lung_mask, get_template, and calculate_DSC from lungct. However, we define convergence as having a Dice similarity coefficient (DSC) between successive iterations of at least 0.99. We recommend using parallelization for the registration step.
95		93
96	![Template Creation](inst/extdata/templatecreation.png)	94	![Template Creation](inst/extdata/templatecreation.png)
97		95
98	``` r	96	```r
99	# Obtain new template	97	# Obtain new template
100	template_new <- get_template(	98	template_new <- get_template(
101	folder_warp = warped_masks,	99	folder_warp = warped_masks,
102	folder_comp = transformations)	100	folder_comp = transformations)
103		101
104	# Check convergence	102	# Check convergence
105	dice <- calculate_DSC(template_init, template_new)	103	dice <- calculate_DSC(template_init, template_new)
106	```	104	```
107		105
108	Radiomics	106	## Radiomics
109	---------	107
110
111	Radiomics, a field of quantitative imaging where large amounts of features are extracted from medical images, is common for lung CTs.	108	Radiomics, a field of quantitative imaging where large amounts of features are extracted from medical images, is common for lung CTs.
112		109
113	Options in lungct:	110	Options in lungct:
114		111
115	- radiomics\_slice: Calculation on 2D slices in axial, coronal, or sagittal planes	112	* radiomics_slice: Calculation on 2D slices in axial, coronal, or sagittal planes
116		113
117	- radiomics\_lung: Calculation on 3D right and left lungs	114	* radiomics_lung: Calculation on 3D right and left lungs
118		115
119	- RIA\_lung: Calculation of advanced radiomic features, such as GLCM and GLRLM [More info](https://github.com/cran/RIA)	116	* RIA_lung: Calculation of advanced radiomic features, such as GLCM and GLRLM [More info](https://github.com/cran/RIA)
120		117
121	``` r	118	```r
122	# Calculate radiomics	119	# Calculate radiomics
123	rad <- RIA_lung(	120	rad <- RIA_lung(
124	img, mask,	121	img, mask,
125	sides = c("right", "left"),	122	sides = c("right", "left"),
126	features = c("fo", "glcm"),	123	features = c("fo", "glcm"),
127	bins_in = 16, equal_prob = FALSE, distance = 1,	124	bins_in = 16, equal_prob = FALSE, distance = 1,
128	statistic = "mean(X, na.rm = TRUE)")	125	statistic = "mean(X, na.rm = TRUE)")
129	```	126	```
130		127
131	COPDGene data usage	128
132	-------------------	129	## COPDGene data usage
133		130
134	For the lung CT template, the [COPDGene data](http://www.copdgene.org/) was used:	131	For the lung CT template, the [COPDGene data](http://www.copdgene.org/) was used:
135		132
136	"This research used data generated by the COPDGene study, which was supported by NIH grants U01HL089856 and U01HL089897. The COPDGene project is also supported by the COPD Foundation through contributions made by an Industry Advisory Board comprised of Pfizer, AstraZeneca, Boehringer Ingelheim, Novartis, and Sunovion."	133	"This research used data generated by the COPDGene study, which was supported by NIH grants U01HL089856 and U01HL089897. The COPDGene project is also supported by the COPD Foundation through contributions made by an Industry Advisory Board comprised of Pfizer, AstraZeneca, Boehringer Ingelheim, Novartis, and Sunovion."
		134