DABC-Net-for-COVID-19 / Git / Diff of /README.md

Models:

RichardZick/

DABC-Net-for-COVID-19

Downloads: 1

Diff of /README.md [8271d0] .. [e67bde]

Switch to unified view


# DABC-Net

DABC-Net toolkit is designed for fast and robust pneumonia segmentation and prediction of COVID-19 progression on chest CT scans. 
The core of the toolkit, DABC-Net, is a novel deep learning (DL) network that combines a 2D U-net for intra-slice spatial information processing, and a recurrent LSTM network to leverage inter-slice context. 
Compared to other popular volumetric segmentation networks such as 3D U-net, DABC-Net is much faster and more robust to CT scans with various slice thickness. 

Based on DABC-Net segmentation, we can predict the disease progression, i.e. whether a specific patient will develop into a severe stage or not using his/her first two CT scans. 
This repository provides an implementation of DABC-Net (including graphical user interface), which can be potentially used to support early triage of severe patients

<b>The main features:</b>
* Ready-to-use (You can run our toolkit with GUI and even no need to install Tensorflow or Python interpreter on your computer.)
* Run everywhere (Desktop app, Web or console)
* Data Anonymization by deleting CT header file
* Fast segmentation
* Built-in multi-types uncertainty
* Prediction of patient progression: mild vs severe
* Support for Covid-19 longitudinal study

## Table of Contents
* [Installation](#installation)
* [Quick start](#quick-start)
    + [DABC-Net for desktop app](#dabc-net-for-desktop-app)
    + [DABC-Net for Colab](#dabc-net-for-colab)  [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb)
    + [DABC-Net for Website](#dabc-net-for-website)
* [Progress prediction](#progress-prediction)
    + [Model](#model)
    + [Usage](#usage)   
    + [Visualization of progress](#visualization-of-progress)
* [Data](#data)
* [Tutorial](#tutorial)

## Installation
If you run this toolkit with packaged desktop app, you can skip this step.

An Nvidia GPU is needed for faster inference (about 16ms/slice on 1080ti gpu).

Requirements:

* tensorflow-gpu == 1.15.4
* Keras == 2.2.4
* numpy == 1.16
* scikit-learn == 0.21.3
* scikit-image == 0.14
* xgboost == 1.1.0
* simpleitk == 2.0
* scipy == 1.1

Install dependencies:

```
cd path/to/repository/
pip install -r requirement.txt
```
The project folder looks like this:

```
path
├─ ... (Codes in DABC-Net-for-COVID-19 repository. Use download.sh to get following files) 
│
├─Input_data
│      2020034797_0123_2949_20200123015940_4.nii.gz
│      2020034797_0125_3052_20200125111145_4.nii.gz
│      ...
│
├─Output_data
│   │
│   ├─covid
│   │      2020034797_0123_2949_20200123015940_4.nii.gz
│   │      2020034797_0125_3052_20200125111145_4.nii.gz
│   │      ...
│   │
│   ├─lung
│   │      2020034797_0123_2949_20200123015940_4.nii.gz
│   │      2020034797_0125_3052_20200125111145_4.nii.gz
│   │      ...
│   │
│   └─uncertainty
│           2020034797_0123_2949_20200123015940_4_predictive_aleatoric.nii.gz
│           2020034797_0125_3052_20200125111145_4_sample_1.nii.gz
│           ...
│
├─weight
│       model_05090017
│       ...
│
│ (following folders are required if you need longitudinal study)
│
├─meta
│       2020035021.csv
│
└─model
        prediction.pkl
        ...

```

## Quick Start

### DABC-Net for desktop app
#### Inference:
1. Download and double click the DABC_Net.exe(Windows) or DABC_Mac(Mac OS) file. 
You can run our network even without installing Tensorflow or Python interpreter on you computer. 
The UI looks like this:



2. Type or select the input folder where you store nii/nii.gz format CT scans data. The output results will be saved in the folder you specified.

3. Choose the sform code name,  the default value is 'NIFTI_XFORM_SCANNER_ANAT'. Some scans without complete header files may loss this value(e.g. data from radiopaedia.org).In this case, please remember select sform name as 'OTHERS'. For more details about header files, please see this [site](https://brainder.org/2012/09/23/the-nifti-file-format/  "With a Title"). 

4. Click 'Run' button. After all the inference done, the progress bar window will be closed. 
   
   ![fig.2](fig/fig2.png)   
   
   Here are some examples:
   ![fig.4](fig/fig4.png )

#### Uncertainty:

In DABC-Net, we approximate Bayersian inference using [DropBlock](http://papers.nips.cc/paper/8271-dropblock-a-regularization-method-for-convolutional-networks), a form of Monte Carlo dropout. For more details about aleatory and epistemic uncertainty, please refer to this [paper](https://pdfs.semanticscholar.org/146f/8844a380191a3f883c3584df3d7a6a56a999.pdf).

1. Follow instructions from item(1-3) in the section above.

2. Choose sample times (integer, e.g. 10). The network will sample 10 times to  compute aleatory/epistemic uncertainty and get mean prediction outcome as final segmentation.

3. Set 'threshold' to get binary output if you need. The default value is 0.5. If you want to save raw probability map from the last sigmoid activation layer of the network, just set threshold  to 0.

4. 'Method' denotes what kind of uncertainty you  want to save.

   ![](fig/fig3.png)

   Here are some examples:

   ![](fig/fig5.png)

#### Visualization:

* Raw: original CT scan
* Lung: output of lung segmentation(optional)
* Lesion: output of lesion segmentation

Then, choose appropriate HU range (e.g. -1024~512) via right slide window.

![](fig/tool_visual.png)

#### Progress predict:
* Meta data: Csv format data. Put the path of data or click 'Demo' button to get an example.
* Method: Use First two scans / First three scans / First scan to predict the progress of disease.
* Output path: (optional) Results will be saved to a text file. If this value is empty, file will save in working directory. 

![](fig/predict_ui.png)


## DABC-Net for Colab

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb)

#### Inference:
1. Put your data in a folder.
2. Select the input and output folder, and run following command:
```
DABC_infer(input_path, output_path, usage, sform_code)
```
- nii_path: 
    - Input: Folder path of input data(nii or nii.gz format).
    - Type: string
- save_path: 
    - Input: Folder path of output data(nii or nii.gz format). The segmentation results will be saved as nii.gz format.
    - Type: string 
- usage:
   - Input: Inference type.
   - Type: string, 'lung' or 'covid'(default)
- sform_code:
   - Input: Coordinate system. In general, some scans without header files(e.g. data from radiopaedia.org) got 'NIFTI_XFORM_UNKNOWN' sform code.
   - Type: bool, 1 for 'NIFTI_XFORM_SCANNER_ANAT'(default) or 0 for 'OTHERS'.
   
#### Uncertainty:
```
DABC_uncertainty(nii_filename, save_filename, sample_value, uncertainty, sform_code)
```
- nii_filename: 
    - Input: Path of input data(nii or nii.gz format).
    - Type: string
- save_filename:
    - Input: Folder path of output data(nii or nii.gz format).
    - Type: string 
- sample_value:
   - Input: number of Monte carlo samples.
   - Type: int
- uncertainty:
   - Input: Choose uncertainty. The results will be saved as nii.gz format.
   - Type: string, 'Predictive','Aleatoric','Epistemic' or 'Both'
- sform_code:
   - Input: Coordinate system. In general, some scans without header files(e.g. data from radiopaedia.org) got 'NIFTI_XFORM_UNKNOWN' sform code.
   - Type: bool, 1 for 'NIFTI_XFORM_SCANNER_ANAT'(default) or 0 for 'OTHERS'.
  

For more detail, please refer to [notebook](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb).

##  DABC-Net for Website


- [ ] Update by Dec 20

# Progress prediction

## Model
### Feature
Feature we used:

| Feature                   | Scan |
| ------------------------- | ------------ |
| Left lesion volume        | scan0 & scan1|
| Left lung volume          | scan0 & scan1|
| Left lesion ratio         | scan0 & scan1|
| Left consolidation volume | scan0 & scan1|
| Left weighted volume      | scan0 & scan1|
| Left z-position           | scan0 & scan1|
| Right lesion volume       | scan0 & scan1|
| Right lung volume         | scan0 & scan1|
| Right lesion ratio        | scan0 & scan1|
| Right consolidation volume| scan0 & scan1|
| Right weighted volume     | scan0 & scan1|
| Right z-position          | scan0 & scan1|
| Age                       | scan0 & scan1|
| Sex                       | scan0 & scan1|

### Base learner
Base learners we used:

| Base learner | MinMaxScaler Necessity | Feature Importance |
| ------------ | ---------------------- | ------------------ |
|SVM            | True  | False |
|MLP            | True  | False |
|Logistic Regression    | True  | False |
|Naive Bayes        | False | False |
|Random Forest      | False | True  |
|Adaboost       | False | True  |
|Gradient Boost     | False | True  |
|XGBoost        | False | True  |

#### MinMaxScalar
For base learners sensitive to data normalization(svm, mlp, ...), we provide the min max normalization based on our training dataset. The weights without min max scalar (TODO) are also provided with fewer base learners and lower performance.

## Usage
### Prediction
```
pred = predict_base_learners(base_learners, feature)
```
- base_learners: 
   - Input: Trained base learners.
   - Type: dict, shape: {key: learner}, key: name of learner, learner: sklearn learner.
- feature: 
   - Input: Preprocessed features.
   - Type: array, shape: m x n, m: number of samples, n: number of features.
- pred: 
   - Output: Probability predicted of base learners. 
   - Type: array, shape: m x k, m: number of samples, k: number of base learners.

## Visualization of progress
Here are some examples:

#### Progression curve of severe patient：

![](fig/progress_curve_severe.png)

#### Progression curve of mild patient：

![](fig/progress_curve_mild.png)

x-axis: time(day), y-axis: lesion ratio

#####  Visualization of different time point scans

![](fig/progress_severe.png)

![](fig/progress_mild.png)

# Data

Dataset with Expert Annotations and Benchmark
* [1] - Ma Jun, Ge Cheng, Wang Yixin, An Xingle, Gao Jiantao, … He Jian. (2020). COVID-19 CT Lung and Infection Segmentation Dataset (Version Verson 1.0) [Data set]. Zenodo. [DOI](https://zenodo.org/record/3757476)

Data Sources
* [2] - Paiva, O., 2020. CORONACASES.ORG - Helping Radiologists To Help People In More Than 100 Countries! \| Coronavirus Cases - 冠状病毒病例. [online] Coronacases.org. Available at: [link](https://Coronacases.org) [Accessed 20 March 2020].
* [3] - Glick, Y., 2020. Viewing Playlist: COVID-19 Pneumonia \| Radiopaedia.Org. [online] Radiopaedia.org. Available at: [link](https://Radiopaedia.org) [Accessed 20 April 2020].

# Notes

Acknowledgements: We thank [COVID-19-CT-Seg-Benchmark repository](https://github.com/JunMa11/COVID-19-CT-Seg-Benchmark) for providing covid-19 segmentation dataset and benchmark. We also thank this [repository](https://github.com/EdwinZhang1970/Python/tree/master/tkinter-pack%20Demo) for providing us ideas for designing ui.

Disclaimer: This toolkit is only for research purpose and not approved for clinical use.


	a/README.md		b/README.md
1	# DABC-Net	1	# DABC-Net
2		2
3	DABC-Net toolkit is designed for fast and robust pneumonia segmentation and prediction of COVID-19 progression on chest CT scans.	3	DABC-Net toolkit is designed for fast and robust pneumonia segmentation and prediction of COVID-19 progression on chest CT scans.
4	The core of the toolkit, DABC-Net, is a novel deep learning (DL) network that combines a 2D U-net for intra-slice spatial information processing, and a recurrent LSTM network to leverage inter-slice context.	4	The core of the toolkit, DABC-Net, is a novel deep learning (DL) network that combines a 2D U-net for intra-slice spatial information processing, and a recurrent LSTM network to leverage inter-slice context.
5	Compared to other popular volumetric segmentation networks such as 3D U-net, DABC-Net is much faster and more robust to CT scans with various slice thickness.	5	Compared to other popular volumetric segmentation networks such as 3D U-net, DABC-Net is much faster and more robust to CT scans with various slice thickness.
6		6
7	Based on DABC-Net segmentation, we can predict the disease progression, i.e. whether a specific patient will develop into a severe stage or not using his/her first two CT scans.	7	Based on DABC-Net segmentation, we can predict the disease progression, i.e. whether a specific patient will develop into a severe stage or not using his/her first two CT scans.
8	This repository provides an implementation of DABC-Net (including graphical user interface), which can be potentially used to support early triage of severe patients	8	This repository provides an implementation of DABC-Net (including graphical user interface), which can be potentially used to support early triage of severe patients
9		9
10	<b>The main features:</b>	10	<b>The main features:</b>
11	* Ready-to-use (You can run our toolkit with GUI and even no need to install Tensorflow or Python interpreter on your computer.)	11	* Ready-to-use (You can run our toolkit with GUI and even no need to install Tensorflow or Python interpreter on your computer.)
12	* Run everywhere (Desktop app, Web or console)	12	* Run everywhere (Desktop app, Web or console)
13	* Data Anonymization by deleting CT header file	13	* Data Anonymization by deleting CT header file
14	* Fast segmentation	14	* Fast segmentation
15	* Built-in multi-types uncertainty	15	* Built-in multi-types uncertainty
16	* Prediction of patient progression: mild vs severe	16	* Prediction of patient progression: mild vs severe
17	* Support for Covid-19 longitudinal study	17	* Support for Covid-19 longitudinal study
18		18
19	## Table of Contents	19	## Table of Contents
20	* [Installation](#installation)	20	* [Installation](#installation)
21	* [Quick start](#quick-start)	21	* [Quick start](#quick-start)
22	+ [DABC-Net for desktop app](#dabc-net-for-desktop-app)	22	+ [DABC-Net for desktop app](#dabc-net-for-desktop-app)
23	+ [DABC-Net for Colab](#dabc-net-for-colab) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb)	23	+ [DABC-Net for Colab](#dabc-net-for-colab) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb)
24	+ [DABC-Net for Website](#dabc-net-for-website)	24	+ [DABC-Net for Website](#dabc-net-for-website)
25	* [Progress prediction](#progress-prediction)	25	* [Progress prediction](#progress-prediction)
26	+ [Model](#model)	26	+ [Model](#model)
27	+ [Usage](#usage)	27	+ [Usage](#usage)
28	+ [Visualization of progress](#visualization-of-progress)	28	+ [Visualization of progress](#visualization-of-progress)
29	* [Data](#data)	29	* [Data](#data)
30	* [Tutorial](#tutorial)	30	* [Tutorial](#tutorial)
31		31
32	## Installation	32	## Installation
33	If you run this toolkit with packaged desktop app, you can skip this step.	33	If you run this toolkit with packaged desktop app, you can skip this step.
34		34
35	An Nvidia GPU is needed for faster inference (about 16ms/slice on 1080ti gpu).	35	An Nvidia GPU is needed for faster inference (about 16ms/slice on 1080ti gpu).
36		36
37	Requirements:	37	Requirements:
38		38
39	* tensorflow-gpu == 1.15.4	39	* tensorflow-gpu == 1.15.4
40	* Keras == 2.2.4	40	* Keras == 2.2.4
41	* numpy == 1.16	41	* numpy == 1.16
42	* scikit-learn == 0.21.3	42	* scikit-learn == 0.21.3
43	* scikit-image == 0.14	43	* scikit-image == 0.14
44	* xgboost == 1.1.0	44	* xgboost == 1.1.0
45	* simpleitk == 2.0	45	* simpleitk == 2.0
46	* scipy == 1.1	46	* scipy == 1.1
47		47
48	Install dependencies:	48	Install dependencies:
49		49
50	```	50	```
51	cd path/to/repository/	51	cd path/to/repository/
52	pip install -r requirement.txt	52	pip install -r requirement.txt
53	```	53	```
54	The project folder looks like this:	54	The project folder looks like this:
55		55
56	```	56	```
57	path	57	path
58	├─ ... (Codes in DABC-Net-for-COVID-19 repository. Use download.sh to get following files)	58	├─ ... (Codes in DABC-Net-for-COVID-19 repository. Use download.sh to get following files)
59	│	59	│
60	├─Input_data	60	├─Input_data
61	│ 2020034797_0123_2949_20200123015940_4.nii.gz	61	│ 2020034797_0123_2949_20200123015940_4.nii.gz
62	│ 2020034797_0125_3052_20200125111145_4.nii.gz	62	│ 2020034797_0125_3052_20200125111145_4.nii.gz
63	│ ...	63	│ ...
64	│	64	│
65	├─Output_data	65	├─Output_data
66	│ │	66	│ │
67	│ ├─covid	67	│ ├─covid
68	│ │ 2020034797_0123_2949_20200123015940_4.nii.gz	68	│ │ 2020034797_0123_2949_20200123015940_4.nii.gz
69	│ │ 2020034797_0125_3052_20200125111145_4.nii.gz	69	│ │ 2020034797_0125_3052_20200125111145_4.nii.gz
70	│ │ ...	70	│ │ ...
71	│ │	71	│ │
72	│ ├─lung	72	│ ├─lung
73	│ │ 2020034797_0123_2949_20200123015940_4.nii.gz	73	│ │ 2020034797_0123_2949_20200123015940_4.nii.gz
74	│ │ 2020034797_0125_3052_20200125111145_4.nii.gz	74	│ │ 2020034797_0125_3052_20200125111145_4.nii.gz
75	│ │ ...	75	│ │ ...
76	│ │	76	│ │
77	│ └─uncertainty	77	│ └─uncertainty
78	│ 2020034797_0123_2949_20200123015940_4_predictive_aleatoric.nii.gz	78	│ 2020034797_0123_2949_20200123015940_4_predictive_aleatoric.nii.gz
79	│ 2020034797_0125_3052_20200125111145_4_sample_1.nii.gz	79	│ 2020034797_0125_3052_20200125111145_4_sample_1.nii.gz
80	│ ...	80	│ ...
81	│	81	│
82	├─weight	82	├─weight
83	│ model_05090017	83	│ model_05090017
84	│ ...	84	│ ...
85	│	85	│
86	│ (following folders are required if you need longitudinal study)	86	│ (following folders are required if you need longitudinal study)
87	│	87	│
88	├─meta	88	├─meta
89	│ 2020035021.csv	89	│ 2020035021.csv
90	│	90	│
91	└─model	91	└─model
92	prediction.pkl	92	prediction.pkl
93	...	93	...
94		94
95	```	95	```
96		96
97	## Quick Start	97	## Quick Start
98		98
99	### DABC-Net for desktop app	99	### DABC-Net for desktop app
100	#### Inference:	100	#### Inference:
101	1. Download and double click the DABC_Net.exe(Windows) or DABC_Mac(Mac OS) file.	101	1. Download and double click the DABC_Net.exe(Windows) or DABC_Mac(Mac OS) file.
102	You can run our network even without installing Tensorflow or Python interpreter on you computer.	102	You can run our network even without installing Tensorflow or Python interpreter on you computer.
103	The UI looks like this:	103	The UI looks like this:
104		104
105	![Alt text](fig/fig1.png "fig.1")	105
106
107	2. Type or select the input folder where you store nii/nii.gz format CT scans data. The output results will be saved in the folder you specified.	106	2. Type or select the input folder where you store nii/nii.gz format CT scans data. The output results will be saved in the folder you specified.
108		107
109	3. Choose the sform code name, the default value is 'NIFTI_XFORM_SCANNER_ANAT'. Some scans without complete header files may loss this value(e.g. data from radiopaedia.org).In this case, please remember select sform name as 'OTHERS'. For more details about header files, please see this [site](https://brainder.org/2012/09/23/the-nifti-file-format/ "With a Title").	108	3. Choose the sform code name, the default value is 'NIFTI_XFORM_SCANNER_ANAT'. Some scans without complete header files may loss this value(e.g. data from radiopaedia.org).In this case, please remember select sform name as 'OTHERS'. For more details about header files, please see this [site](https://brainder.org/2012/09/23/the-nifti-file-format/ "With a Title").
110		109
111	4. Click 'Run' button. After all the inference done, the progress bar window will be closed.	110	4. Click 'Run' button. After all the inference done, the progress bar window will be closed.
112		111
113	![fig.2](fig/fig2.png)	112	![fig.2](fig/fig2.png)
114		113
115	Here are some examples:	114	Here are some examples:
116	![fig.4](fig/fig4.png )	115	![fig.4](fig/fig4.png )
117		116
118	#### Uncertainty:	117	#### Uncertainty:
119		118
120	In DABC-Net, we approximate Bayersian inference using [DropBlock](http://papers.nips.cc/paper/8271-dropblock-a-regularization-method-for-convolutional-networks), a form of Monte Carlo dropout. For more details about aleatory and epistemic uncertainty, please refer to this [paper](https://pdfs.semanticscholar.org/146f/8844a380191a3f883c3584df3d7a6a56a999.pdf).	119	In DABC-Net, we approximate Bayersian inference using [DropBlock](http://papers.nips.cc/paper/8271-dropblock-a-regularization-method-for-convolutional-networks), a form of Monte Carlo dropout. For more details about aleatory and epistemic uncertainty, please refer to this [paper](https://pdfs.semanticscholar.org/146f/8844a380191a3f883c3584df3d7a6a56a999.pdf).
121		120
122	1. Follow instructions from item(1-3) in the section above.	121	1. Follow instructions from item(1-3) in the section above.
123		122
124	2. Choose sample times (integer, e.g. 10). The network will sample 10 times to compute aleatory/epistemic uncertainty and get mean prediction outcome as final segmentation.	123	2. Choose sample times (integer, e.g. 10). The network will sample 10 times to compute aleatory/epistemic uncertainty and get mean prediction outcome as final segmentation.
125		124
126	3. Set 'threshold' to get binary output if you need. The default value is 0.5. If you want to save raw probability map from the last sigmoid activation layer of the network, just set threshold to 0.	125	3. Set 'threshold' to get binary output if you need. The default value is 0.5. If you want to save raw probability map from the last sigmoid activation layer of the network, just set threshold to 0.
127		126
128	4. 'Method' denotes what kind of uncertainty you want to save.	127	4. 'Method' denotes what kind of uncertainty you want to save.
129		128
130	![](fig/fig3.png)	129	![](fig/fig3.png)
131		130
132	Here are some examples:	131	Here are some examples:
133		132
134	![](fig/fig5.png)	133	![](fig/fig5.png)
135		134
136	#### Visualization:	135	#### Visualization:
137		136
138	* Raw: original CT scan	137	* Raw: original CT scan
139	* Lung: output of lung segmentation(optional)	138	* Lung: output of lung segmentation(optional)
140	* Lesion: output of lesion segmentation	139	* Lesion: output of lesion segmentation
141		140
142	Then, choose appropriate HU range (e.g. -1024~512) via right slide window.	141	Then, choose appropriate HU range (e.g. -1024~512) via right slide window.
143		142
144	![](fig/tool_visual.png)	143	![](fig/tool_visual.png)
145		144
146	#### Progress predict:	145	#### Progress predict:
147	* Meta data: Csv format data. Put the path of data or click 'Demo' button to get an example.	146	* Meta data: Csv format data. Put the path of data or click 'Demo' button to get an example.
148	* Method: Use First two scans / First three scans / First scan to predict the progress of disease.	147	* Method: Use First two scans / First three scans / First scan to predict the progress of disease.
149	* Output path: (optional) Results will be saved to a text file. If this value is empty, file will save in working directory.	148	* Output path: (optional) Results will be saved to a text file. If this value is empty, file will save in working directory.
150		149
151	![](fig/predict_ui.png)	150	![](fig/predict_ui.png)
152		151
153		152
154	## DABC-Net for Colab	153	## DABC-Net for Colab
155		154
156	[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb)	155	[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb)
157		156
158	#### Inference:	157	#### Inference:
159	1. Put your data in a folder.	158	1. Put your data in a folder.
160	2. Select the input and output folder, and run following command:	159	2. Select the input and output folder, and run following command:
161	```	160	```
162	DABC_infer(input_path, output_path, usage, sform_code)	161	DABC_infer(input_path, output_path, usage, sform_code)
163	```	162	```
164	- nii_path:	163	- nii_path:
165	- Input: Folder path of input data(nii or nii.gz format).	164	- Input: Folder path of input data(nii or nii.gz format).
166	- Type: string	165	- Type: string
167	- save_path:	166	- save_path:
168	- Input: Folder path of output data(nii or nii.gz format). The segmentation results will be saved as nii.gz format.	167	- Input: Folder path of output data(nii or nii.gz format). The segmentation results will be saved as nii.gz format.
169	- Type: string	168	- Type: string
170	- usage:	169	- usage:
171	- Input: Inference type.	170	- Input: Inference type.
172	- Type: string, 'lung' or 'covid'(default)	171	- Type: string, 'lung' or 'covid'(default)
173	- sform_code:	172	- sform_code:
174	- Input: Coordinate system. In general, some scans without header files(e.g. data from radiopaedia.org) got 'NIFTI_XFORM_UNKNOWN' sform code.	173	- Input: Coordinate system. In general, some scans without header files(e.g. data from radiopaedia.org) got 'NIFTI_XFORM_UNKNOWN' sform code.
175	- Type: bool, 1 for 'NIFTI_XFORM_SCANNER_ANAT'(default) or 0 for 'OTHERS'.	174	- Type: bool, 1 for 'NIFTI_XFORM_SCANNER_ANAT'(default) or 0 for 'OTHERS'.
176		175
177	#### Uncertainty:	176	#### Uncertainty:
178	```	177	```
179	DABC_uncertainty(nii_filename, save_filename, sample_value, uncertainty, sform_code)	178	DABC_uncertainty(nii_filename, save_filename, sample_value, uncertainty, sform_code)
180	```	179	```
181	- nii_filename:	180	- nii_filename:
182	- Input: Path of input data(nii or nii.gz format).	181	- Input: Path of input data(nii or nii.gz format).
183	- Type: string	182	- Type: string
184	- save_filename:	183	- save_filename:
185	- Input: Folder path of output data(nii or nii.gz format).	184	- Input: Folder path of output data(nii or nii.gz format).
186	- Type: string	185	- Type: string
187	- sample_value:	186	- sample_value:
188	- Input: number of Monte carlo samples.	187	- Input: number of Monte carlo samples.
189	- Type: int	188	- Type: int
190	- uncertainty:	189	- uncertainty:
191	- Input: Choose uncertainty. The results will be saved as nii.gz format.	190	- Input: Choose uncertainty. The results will be saved as nii.gz format.
192	- Type: string, 'Predictive','Aleatoric','Epistemic' or 'Both'	191	- Type: string, 'Predictive','Aleatoric','Epistemic' or 'Both'
193	- sform_code:	192	- sform_code:
194	- Input: Coordinate system. In general, some scans without header files(e.g. data from radiopaedia.org) got 'NIFTI_XFORM_UNKNOWN' sform code.	193	- Input: Coordinate system. In general, some scans without header files(e.g. data from radiopaedia.org) got 'NIFTI_XFORM_UNKNOWN' sform code.
195	- Type: bool, 1 for 'NIFTI_XFORM_SCANNER_ANAT'(default) or 0 for 'OTHERS'.	194	- Type: bool, 1 for 'NIFTI_XFORM_SCANNER_ANAT'(default) or 0 for 'OTHERS'.
196		195
197		196
198	For more detail, please refer to [notebook](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb).	197	For more detail, please refer to [notebook](https://colab.research.google.com/github/Robin970822/DABC-Net-for-COVID-19/blob/master/DABC_pipeline_demo.ipynb).
199		198
200	## DABC-Net for Website	199	## DABC-Net for Website
201		200
202		201
203	- [ ] Update by Dec 20	202	- [ ] Update by Dec 20
204		203
205	# Progress prediction	204	# Progress prediction
206		205
207	## Model	206	## Model
208	### Feature	207	### Feature
209	Feature we used:	208	Feature we used:
210		209
211	\| Feature \| Scan \|	210	\| Feature \| Scan \|
212	\| ------------------------- \| ------------ \|	211	\| ------------------------- \| ------------ \|
213	\| Left lesion volume \| scan0 & scan1\|	212	\| Left lesion volume \| scan0 & scan1\|
214	\| Left lung volume \| scan0 & scan1\|	213	\| Left lung volume \| scan0 & scan1\|
215	\| Left lesion ratio \| scan0 & scan1\|	214	\| Left lesion ratio \| scan0 & scan1\|
216	\| Left consolidation volume \| scan0 & scan1\|	215	\| Left consolidation volume \| scan0 & scan1\|
217	\| Left weighted volume \| scan0 & scan1\|	216	\| Left weighted volume \| scan0 & scan1\|
218	\| Left z-position \| scan0 & scan1\|	217	\| Left z-position \| scan0 & scan1\|
219	\| Right lesion volume \| scan0 & scan1\|	218	\| Right lesion volume \| scan0 & scan1\|
220	\| Right lung volume \| scan0 & scan1\|	219	\| Right lung volume \| scan0 & scan1\|
221	\| Right lesion ratio \| scan0 & scan1\|	220	\| Right lesion ratio \| scan0 & scan1\|
222	\| Right consolidation volume\| scan0 & scan1\|	221	\| Right consolidation volume\| scan0 & scan1\|
223	\| Right weighted volume \| scan0 & scan1\|	222	\| Right weighted volume \| scan0 & scan1\|
224	\| Right z-position \| scan0 & scan1\|	223	\| Right z-position \| scan0 & scan1\|
225	\| Age \| scan0 & scan1\|	224	\| Age \| scan0 & scan1\|
226	\| Sex \| scan0 & scan1\|	225	\| Sex \| scan0 & scan1\|
227		226
228	### Base learner	227	### Base learner
229	Base learners we used:	228	Base learners we used:
230		229
231	\| Base learner \| MinMaxScaler Necessity \| Feature Importance \|	230	\| Base learner \| MinMaxScaler Necessity \| Feature Importance \|
232	\| ------------ \| ---------------------- \| ------------------ \|	231	\| ------------ \| ---------------------- \| ------------------ \|
233	\|SVM \| True \| False \|	232	\|SVM \| True \| False \|
234	\|MLP \| True \| False \|	233	\|MLP \| True \| False \|
235	\|Logistic Regression \| True \| False \|	234	\|Logistic Regression \| True \| False \|
236	\|Naive Bayes \| False \| False \|	235	\|Naive Bayes \| False \| False \|
237	\|Random Forest \| False \| True \|	236	\|Random Forest \| False \| True \|
238	\|Adaboost \| False \| True \|	237	\|Adaboost \| False \| True \|
239	\|Gradient Boost \| False \| True \|	238	\|Gradient Boost \| False \| True \|
240	\|XGBoost \| False \| True \|	239	\|XGBoost \| False \| True \|
241		240
242	#### MinMaxScalar	241	#### MinMaxScalar
243	For base learners sensitive to data normalization(svm, mlp, ...), we provide the min max normalization based on our training dataset. The weights without min max scalar (TODO) are also provided with fewer base learners and lower performance.	242	For base learners sensitive to data normalization(svm, mlp, ...), we provide the min max normalization based on our training dataset. The weights without min max scalar (TODO) are also provided with fewer base learners and lower performance.
244		243
245	## Usage	244	## Usage
246	### Prediction	245	### Prediction
247	```	246	```
248	pred = predict_base_learners(base_learners, feature)	247	pred = predict_base_learners(base_learners, feature)
249	```	248	```
250	- base_learners:	249	- base_learners:
251	- Input: Trained base learners.	250	- Input: Trained base learners.
252	- Type: dict, shape: {key: learner}, key: name of learner, learner: sklearn learner.	251	- Type: dict, shape: {key: learner}, key: name of learner, learner: sklearn learner.
253	- feature:	252	- feature:
254	- Input: Preprocessed features.	253	- Input: Preprocessed features.
255	- Type: array, shape: m x n, m: number of samples, n: number of features.	254	- Type: array, shape: m x n, m: number of samples, n: number of features.
256	- pred:	255	- pred:
257	- Output: Probability predicted of base learners.	256	- Output: Probability predicted of base learners.
258	- Type: array, shape: m x k, m: number of samples, k: number of base learners.	257	- Type: array, shape: m x k, m: number of samples, k: number of base learners.
259		258
260	## Visualization of progress	259	## Visualization of progress
261	Here are some examples:	260	Here are some examples:
262		261
263	#### Progression curve of severe patient：	262	#### Progression curve of severe patient：
264		263
265	![](fig/progress_curve_severe.png)	264	![](fig/progress_curve_severe.png)
266		265
267	#### Progression curve of mild patient：	266	#### Progression curve of mild patient：
268		267
269	![](fig/progress_curve_mild.png)	268	![](fig/progress_curve_mild.png)
270		269
271	x-axis: time(day), y-axis: lesion ratio	270	x-axis: time(day), y-axis: lesion ratio
272		271
273	##### Visualization of different time point scans	272	##### Visualization of different time point scans
274		273
275	![](fig/progress_severe.png)	274	![](fig/progress_severe.png)
276		275
277	![](fig/progress_mild.png)	276	![](fig/progress_mild.png)
278		277
279	# Data	278	# Data
280		279
281	Dataset with Expert Annotations and Benchmark	280	Dataset with Expert Annotations and Benchmark
282	* [1] - Ma Jun, Ge Cheng, Wang Yixin, An Xingle, Gao Jiantao, … He Jian. (2020). COVID-19 CT Lung and Infection Segmentation Dataset (Version Verson 1.0) [Data set]. Zenodo. [DOI](https://zenodo.org/record/3757476)	281	* [1] - Ma Jun, Ge Cheng, Wang Yixin, An Xingle, Gao Jiantao, … He Jian. (2020). COVID-19 CT Lung and Infection Segmentation Dataset (Version Verson 1.0) [Data set]. Zenodo. [DOI](https://zenodo.org/record/3757476)
283		282
284	Data Sources	283	Data Sources
285	* [2] - Paiva, O., 2020. CORONACASES.ORG - Helping Radiologists To Help People In More Than 100 Countries! \\| Coronavirus Cases - 冠状病毒病例. [online] Coronacases.org. Available at: [link](https://Coronacases.org) [Accessed 20 March 2020].	284	* [2] - Paiva, O., 2020. CORONACASES.ORG - Helping Radiologists To Help People In More Than 100 Countries! \\| Coronavirus Cases - 冠状病毒病例. [online] Coronacases.org. Available at: [link](https://Coronacases.org) [Accessed 20 March 2020].
286	* [3] - Glick, Y., 2020. Viewing Playlist: COVID-19 Pneumonia \\| Radiopaedia.Org. [online] Radiopaedia.org. Available at: [link](https://Radiopaedia.org) [Accessed 20 April 2020].	285	* [3] - Glick, Y., 2020. Viewing Playlist: COVID-19 Pneumonia \\| Radiopaedia.Org. [online] Radiopaedia.org. Available at: [link](https://Radiopaedia.org) [Accessed 20 April 2020].
287		286
288	# Notes	287	# Notes
289		288
290	Acknowledgements: We thank [COVID-19-CT-Seg-Benchmark repository](https://github.com/JunMa11/COVID-19-CT-Seg-Benchmark) for providing covid-19 segmentation dataset and benchmark. We also thank this [repository](https://github.com/EdwinZhang1970/Python/tree/master/tkinter-pack%20Demo) for providing us ideas for designing ui.	289	Acknowledgements: We thank [COVID-19-CT-Seg-Benchmark repository](https://github.com/JunMa11/COVID-19-CT-Seg-Benchmark) for providing covid-19 segmentation dataset and benchmark. We also thank this [repository](https://github.com/EdwinZhang1970/Python/tree/master/tkinter-pack%20Demo) for providing us ideas for designing ui.
291		290
292	Disclaimer: This toolkit is only for research purpose and not approved for clinical use.	291	Disclaimer: This toolkit is only for research purpose and not approved for clinical use.
293		292