|
a |
|
b/README.md |
|
|
1 |
# Brain MRI Segmentation |
|
|
2 |
|
|
|
3 |
The method we use comes from this paper: |
|
|
4 |
[From neonatal to adult brain |
|
|
5 |
mr image segmentation in a few seconds using 3d-like fully convolutional network and transfer learning](https://www.lrde.epita.fr/wiki/Publications/xu.17.icip) |
|
|
6 |
|
|
|
7 |
Soft tissue segmentation. |
|
|
8 |
|
|
|
9 |
This project was part of the Smart India Hackathon 2019 in which our team was the runner ups. |
|
|
10 |
The problem statement was **Brain MRI Segmentation using Machine Learning** given by **Department of Atomic Energy, Government of India** |
|
|
11 |
|
|
|
12 |
This project could be used by medical professionals for medical purposes. |
|
|
13 |
|
|
|
14 |
 |
|
|
15 |
|
|
|
16 |
## Preprocessing |
|
|
17 |
|
|
|
18 |
 |
|
|
19 |
|
|
|
20 |
|
|
|
21 |
 |
|
|
22 |
|
|
|
23 |
 |
|
|
24 |
|
|
|
25 |
## Why our model? |
|
|
26 |
|
|
|
27 |
• Since we are using transfer learning, a novel approach in this field, so we do not need to train our model from scratch which makes it very fast in training in comparison to other models. |
|
|
28 |
|
|
|
29 |
• Stacking 3 successive 2D slices allows us to make a RGB image, another novel idea.This representation enables us to incorporate some 3D information, while avoiding the expensive computational and memory requirements of fully 3D FCN. |
|
|
30 |
|
|
|
31 |
• Using Transfer Learning we do not need many training images, so we could train our model very well only on a few training images. |
|
|
32 |
|
|
|
33 |
• We are also using traditional data augmentation methods like rotating, cropping and flipping the images in training set for improving our model. |
|
|
34 |
|
|
|
35 |
## GUI and giving input |
|
|
36 |
|
|
|
37 |
 |
|
|
38 |
|
|
|
39 |
|
|
|
40 |
## Output |
|
|
41 |
|
|
|
42 |
 |
|
|
43 |
|
|
|
44 |
## Tumour prediction |
|
|
45 |
|
|
|
46 |
 |
|
|
47 |
|
|
|
48 |
## Other regions |
|
|
49 |
|
|
|
50 |
 |
|
|
51 |
|
|
|
52 |
## Contributors |
|
|
53 |
[Vaibhav Shukla](https://github.com/vaibhavshukla182/) |
|
|
54 |
[Abhijeet Singh](https://github.com/abhi40308) |
|
|
55 |
[Omkar Ajnadkar](https://github.com/blackbird71SR) |
|
|
56 |
[Govind Singh Rajpurohit](https://github.com/govind51) |
|
|
57 |
[Ratna Priya](https://github.com/Ratna04priya) |
|
|
58 |
[Sanath Singavarapu](https://github.com/Killer2499) |