SA-MIL / Git / Diff of /README.md

Models:

DavidFeaster/

SA-MIL

Downloads: 1

Diff of /README.md [05c4fa] .. [72fbc7]

Switch to unified view


<!-- Improved compatibility of back to top link: See: https://github.com/othneildrew/Best-README-Template/pull/73 -->
<a name="readme-top"></a>
<!--
*** Thanks for checking out the Best-README-Template. If you have a suggestion
*** that would make this better, please fork the repo and create a pull request
*** or simply open an issue with the tag "enhancement".
*** Don't forget to give the project a star!
*** Thanks again! Now go create something AMAZING! :D
-->



<!-- PROJECT SHIELDS -->
<!--
*** I'm using markdown "reference style" links for readability.
*** Reference links are enclosed in brackets [ ] instead of parentheses ( ).
*** See the bottom of this document for the declaration of the reference variables
*** for contributors-url, forks-url, etc. This is an optional, concise syntax you may use.
*** https://www.markdownguide.org/basic-syntax/#reference-style-links
-->








  <h3 align="center">Smooth Attention for Deep Multiple Instance
Learning: Application to CT Intracranial
Hemorrhage Detection</h3>

  <p align="center">
    The codes for the paper accepted in MICCAI 2023.
  </p>
</div>



<!-- TABLE OF CONTENTS -->
<details>
  <summary>Table of Contents</summary>
  <ol>
    <li>
      <a href="#introduction">Introduction</a>
    </li>
    <li>
      <a href="#getting-started">Getting Started</a>
      <ul>
        <li><a href="#prerequisites">Prerequisites</a></li>
        <li><a href="#installation">Installation</a></li>
      </ul>
    </li>
    <li><a href="#usage">Usage</a></li>
  </ol>
</details>



<!-- INTRODUCTION -->
## Introduction

Multiple Instance Learning (MIL) has been widely applied to medical imaging diagnosis, where bag labels are known and instance labels inside bags are unknown. Traditional MIL assumes that instances in each bag are independent samples from a given distribution. However, instances are often spatially or sequentially ordered, and one would expect similar diagnostic importance for neighboring instances. To address this, in this study, we propose a smooth attention deep MIL (SA-DMIL) model. Smoothness is achieved by the introduction of first and second order constraints on the latent function encoding the attention paid to each instance in a bag. The method is applied to the detection of intracranial hemorrhage (ICH) on head CT scans.
The results show that this novel SA-DMIL: (a) achieves better performance than the non-smooth attention MIL at both scan (bag) and slice (instance) levels; (b) learns spatial dependencies between slices; and (c) outperforms current state-of-the-art MIL methods on the same ICH test set.   

<p align="right">(<a href="#readme-top">back to top</a>)</p>


<!-- GETTING STARTED -->
## Getting Started

This is an example of how you may give instructions on setting up your project locally.
To get a local copy up and running follow these simple example steps.

### Prerequisites

This is an example of how to list things you need to use the software and how to install them.
* The codes use Tensorflow and you can download all packages in [requirements.txt](https://github.com/YunanWu2168/SA-MIL/blob/master/requirements.txt).

```
matplotlib==3.7.1

numpy==1.22.4

opencv_contrib_python==4.7.0.72

opencv_python==4.7.0.72

opencv_python_headless==4.7.0.72

pandas==1.5.3

scikit_learn==1.2.2

tensorflow==2.12.0
```

### Installation

1. Pip install [requirements.txt](https://github.com/YunanWu2168/SA-MIL/blob/master/requirements.txt)
   ```sh
   pip install requirements.txt
   ```
2. Open SA-MIL-preprocessing.ipynb - How to process head CTs
3. Open SA-MIL-training.ipynb - Train SA-DMIL
4. Open Non-SA-MIL-training.ipynb - Train Att-MIL
5. Open SA-MIL-testing.ipynb - Test SA-DMIL and Att-MIL
6. Open vis-SA-MIL.ipynb - Visualize at slice level

<p align="right">(<a href="#readme-top">back to top</a>)</p>

## Dataset

### Download
The dataset used in this paper can be download via [Kaggle Challenge Dataset](https://www.kaggle.com/competitions/rsna-intracranial-hemorrhage-detection/data)

### CT Slice Image Processing with Windowing 
[SA_MIL_preprocessing.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_preprocessing.ipynb)

<!-- USAGE EXAMPLES -->
## Usage

1. Model training at scan level

  [SA_MIL_training.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_training.ipynb)

  [Non_SA_MIL_training.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/Non_SA_MIL_training.ipynb)

2. Model testing at scan level

  [SA_MIL_testing.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_testing.ipynb)

3. Model testing at slice level

  [vis_SA_MIL.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/vis_SA_MIL.ipynb)

<p align="right">(<a href="#readme-top">back to top</a>)</p>


<!-- LICENSE -->
## License

Distributed under the MIT License. See `LICENSE.txt` for more information.

<p align="right">(<a href="#readme-top">back to top</a>)</p>


	a/README.md		b/README.md
1	<!-- Improved compatibility of back to top link: See: https://github.com/othneildrew/Best-README-Template/pull/73 -->	1	<!-- Improved compatibility of back to top link: See: https://github.com/othneildrew/Best-README-Template/pull/73 -->
2	<a name="readme-top"></a>	2	<a name="readme-top"></a>
3	<!--	3	<!--
4	*** Thanks for checking out the Best-README-Template. If you have a suggestion	4	*** Thanks for checking out the Best-README-Template. If you have a suggestion
5	*** that would make this better, please fork the repo and create a pull request	5	*** that would make this better, please fork the repo and create a pull request
6	*** or simply open an issue with the tag "enhancement".	6	*** or simply open an issue with the tag "enhancement".
7	*** Don't forget to give the project a star!	7	*** Don't forget to give the project a star!
8	*** Thanks again! Now go create something AMAZING! :D	8	*** Thanks again! Now go create something AMAZING! :D
9	-->	9	-->
10		10
11		11
12		12
13	<!-- PROJECT SHIELDS -->	13	<!-- PROJECT SHIELDS -->
14	<!--	14	<!--
15	*** I'm using markdown "reference style" links for readability.	15	*** I'm using markdown "reference style" links for readability.
16	*** Reference links are enclosed in brackets [ ] instead of parentheses ( ).	16	*** Reference links are enclosed in brackets [ ] instead of parentheses ( ).
17	*** See the bottom of this document for the declaration of the reference variables	17	*** See the bottom of this document for the declaration of the reference variables
18	*** for contributors-url, forks-url, etc. This is an optional, concise syntax you may use.	18	*** for contributors-url, forks-url, etc. This is an optional, concise syntax you may use.
19	*** https://www.markdownguide.org/basic-syntax/#reference-style-links	19	*** https://www.markdownguide.org/basic-syntax/#reference-style-links
20	-->	20	-->
21		21
22	<!-- PROJECT LOGO -->	22
23	<br />	23
24	<div align="center">
25	<a href="https://github.com/othneildrew/Best-README-Template">
26	<img src="images/logo.png" alt="Logo" width="80" height="80">
27	</a>
28
29	<h3 align="center">Smooth Attention for Deep Multiple Instance	24	<h3 align="center">Smooth Attention for Deep Multiple Instance
30	Learning: Application to CT Intracranial	25	Learning: Application to CT Intracranial
31	Hemorrhage Detection</h3>	26	Hemorrhage Detection</h3>
32		27
33	<p align="center">	28	<p align="center">
34	The codes for the paper accepted in MICCAI 2023.	29	The codes for the paper accepted in MICCAI 2023.
35	</p>	30	</p>
36	</div>	31	</div>
37		32
38		33
39		34
40	<!-- TABLE OF CONTENTS -->	35	<!-- TABLE OF CONTENTS -->
41	<details>	36	<details>
42	<summary>Table of Contents</summary>	37	<summary>Table of Contents</summary>
43	<ol>	38	<ol>
44	<li>	39	<li>
45	<a href="#introduction">Introduction</a>	40	<a href="#introduction">Introduction</a>
46	</li>	41	</li>
47	<li>	42	<li>
48	<a href="#getting-started">Getting Started</a>	43	<a href="#getting-started">Getting Started</a>
49	<ul>	44	<ul>
50	<li><a href="#prerequisites">Prerequisites</a></li>	45	<li><a href="#prerequisites">Prerequisites</a></li>
51	<li><a href="#installation">Installation</a></li>	46	<li><a href="#installation">Installation</a></li>
52	</ul>	47	</ul>
53	</li>	48	</li>
54	<li><a href="#usage">Usage</a></li>	49	<li><a href="#usage">Usage</a></li>
55	</ol>	50	</ol>
56	</details>	51	</details>
57		52
58		53
59		54
60	<!-- INTRODUCTION -->	55	<!-- INTRODUCTION -->
61	## Introduction	56	## Introduction
62		57
63	Multiple Instance Learning (MIL) has been widely applied to medical imaging diagnosis, where bag labels are known and instance labels inside bags are unknown. Traditional MIL assumes that instances in each bag are independent samples from a given distribution. However, instances are often spatially or sequentially ordered, and one would expect similar diagnostic importance for neighboring instances. To address this, in this study, we propose a smooth attention deep MIL (SA-DMIL) model. Smoothness is achieved by the introduction of first and second order constraints on the latent function encoding the attention paid to each instance in a bag. The method is applied to the detection of intracranial hemorrhage (ICH) on head CT scans.	58	Multiple Instance Learning (MIL) has been widely applied to medical imaging diagnosis, where bag labels are known and instance labels inside bags are unknown. Traditional MIL assumes that instances in each bag are independent samples from a given distribution. However, instances are often spatially or sequentially ordered, and one would expect similar diagnostic importance for neighboring instances. To address this, in this study, we propose a smooth attention deep MIL (SA-DMIL) model. Smoothness is achieved by the introduction of first and second order constraints on the latent function encoding the attention paid to each instance in a bag. The method is applied to the detection of intracranial hemorrhage (ICH) on head CT scans.
64	The results show that this novel SA-DMIL: (a) achieves better performance than the non-smooth attention MIL at both scan (bag) and slice (instance) levels; (b) learns spatial dependencies between slices; and (c) outperforms current state-of-the-art MIL methods on the same ICH test set.	59	The results show that this novel SA-DMIL: (a) achieves better performance than the non-smooth attention MIL at both scan (bag) and slice (instance) levels; (b) learns spatial dependencies between slices; and (c) outperforms current state-of-the-art MIL methods on the same ICH test set.
65		60
66	<p align="right">(<a href="#readme-top">back to top</a>)</p>	61	<p align="right">(<a href="#readme-top">back to top</a>)</p>
67		62
68		63
69	<!-- GETTING STARTED -->	64	<!-- GETTING STARTED -->
70	## Getting Started	65	## Getting Started
71		66
72	This is an example of how you may give instructions on setting up your project locally.	67	This is an example of how you may give instructions on setting up your project locally.
73	To get a local copy up and running follow these simple example steps.	68	To get a local copy up and running follow these simple example steps.
74		69
75	### Prerequisites	70	### Prerequisites
76		71
77	This is an example of how to list things you need to use the software and how to install them.	72	This is an example of how to list things you need to use the software and how to install them.
78	* The codes use Tensorflow and you can download all packages in [requirements.txt](https://github.com/YunanWu2168/SA-MIL/blob/master/requirements.txt).	73	* The codes use Tensorflow and you can download all packages in [requirements.txt](https://github.com/YunanWu2168/SA-MIL/blob/master/requirements.txt).
79		74
80	```	75	```
81	matplotlib==3.7.1	76	matplotlib==3.7.1
82		77
83	numpy==1.22.4	78	numpy==1.22.4
84		79
85	opencv_contrib_python==4.7.0.72	80	opencv_contrib_python==4.7.0.72
86		81
87	opencv_python==4.7.0.72	82	opencv_python==4.7.0.72
88		83
89	opencv_python_headless==4.7.0.72	84	opencv_python_headless==4.7.0.72
90		85
91	pandas==1.5.3	86	pandas==1.5.3
92		87
93	scikit_learn==1.2.2	88	scikit_learn==1.2.2
94		89
95	tensorflow==2.12.0	90	tensorflow==2.12.0
96	```	91	```
97		92
98	### Installation	93	### Installation
99		94
100	1. Pip install [requirements.txt](https://github.com/YunanWu2168/SA-MIL/blob/master/requirements.txt)	95	1. Pip install [requirements.txt](https://github.com/YunanWu2168/SA-MIL/blob/master/requirements.txt)
101	```sh	96	```sh
102	pip install requirements.txt	97	pip install requirements.txt
103	```	98	```
104	2. Open SA-MIL-preprocessing.ipynb - How to process head CTs	99	2. Open SA-MIL-preprocessing.ipynb - How to process head CTs
105	3. Open SA-MIL-training.ipynb - Train SA-DMIL	100	3. Open SA-MIL-training.ipynb - Train SA-DMIL
106	4. Open Non-SA-MIL-training.ipynb - Train Att-MIL	101	4. Open Non-SA-MIL-training.ipynb - Train Att-MIL
107	5. Open SA-MIL-testing.ipynb - Test SA-DMIL and Att-MIL	102	5. Open SA-MIL-testing.ipynb - Test SA-DMIL and Att-MIL
108	6. Open vis-SA-MIL.ipynb - Visualize at slice level	103	6. Open vis-SA-MIL.ipynb - Visualize at slice level
109		104
110	<p align="right">(<a href="#readme-top">back to top</a>)</p>	105	<p align="right">(<a href="#readme-top">back to top</a>)</p>
111		106
112	## Dataset	107	## Dataset
113		108
114	### Download	109	### Download
115	The dataset used in this paper can be download via [Kaggle Challenge Dataset](https://www.kaggle.com/competitions/rsna-intracranial-hemorrhage-detection/data)	110	The dataset used in this paper can be download via [Kaggle Challenge Dataset](https://www.kaggle.com/competitions/rsna-intracranial-hemorrhage-detection/data)
116		111
117	### CT Slice Image Processing with Windowing	112	### CT Slice Image Processing with Windowing
118	[SA_MIL_preprocessing.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_preprocessing.ipynb)	113	[SA_MIL_preprocessing.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_preprocessing.ipynb)
119		114
120	<!-- USAGE EXAMPLES -->	115	<!-- USAGE EXAMPLES -->
121	## Usage	116	## Usage
122		117
123	1. Model training at scan level	118	1. Model training at scan level
124		119
125	[SA_MIL_training.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_training.ipynb)	120	[SA_MIL_training.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_training.ipynb)
126		121
127	[Non_SA_MIL_training.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/Non_SA_MIL_training.ipynb)	122	[Non_SA_MIL_training.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/Non_SA_MIL_training.ipynb)
128		123
129	2. Model testing at scan level	124	2. Model testing at scan level
130		125
131	[SA_MIL_testing.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_testing.ipynb)	126	[SA_MIL_testing.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/SA_MIL_testing.ipynb)
132		127
133	3. Model testing at slice level	128	3. Model testing at slice level
134		129
135	[vis_SA_MIL.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/vis_SA_MIL.ipynb)	130	[vis_SA_MIL.ipynb](https://github.com/YunanWu2168/SA-MIL/blob/master/vis_SA_MIL.ipynb)
136		131
137	<p align="right">(<a href="#readme-top">back to top</a>)</p>	132	<p align="right">(<a href="#readme-top">back to top</a>)</p>
138		133
139		134
140	<!-- LICENSE -->	135	<!-- LICENSE -->
141	## License	136	## License
142		137
143	Distributed under the MIT License. See `LICENSE.txt` for more information.	138	Distributed under the MIT License. See `LICENSE.txt` for more information.
144		139
145	<p align="right">(<a href="#readme-top">back to top</a>)</p>	140	<p align="right">(<a href="#readme-top">back to top</a>)</p>
146		141