Hugging Face's logo

Kaynaaf
/
BrainMRI-Classifier

Keras
medical
healthcare
Model card Files
xet
Community
BrainMRI-Classifier
File size: 2,485 Bytes 
7183b27
 
 
 
 
 
 
 
 
 
 
 
 
582038e
7183b27
 
 
 
d5505d6
7183b27
19dd68e
 
 
 
 
7183b27
 
 
 
 
d5505d6
7183b27
 
 
582038e
7183b27
d5505d6
7183b27
 
 
582038e
7183b27
22e905d
 
 
20b8a25
7183b27
 
 
582038e
7183b27
18ae1b8
 
 
 
 
 
 
 
 
7183b27
72f160c
7183b27
147b396
58ed971
9c7467e
e831d36
5ac51ca
7183b27
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
 
---
license: gpl-3.0
datasets:
- Kaynaaf/Brain-Tumour-MRI
metrics:
- accuracy 0.90
- precision 0.90
library_name: keras
tags:
- medical
- healthcare
---

# Model Card 

An Image Classifier that predicts the presence of certain Brain tumours from their MRI scans

## Model Details
A 134M Parameter ConvNet designed for classification of Brain tumours in MRI scans.

## Paper

Interpretable Deep Learning for Brain Tumor Diagnosis: Occlusion Sensitivity-Driven Explainability in MRI Classification
DOI: [10.21015/vtse.v13i2.2082](10.21015/vtse.v13i2.2082)

## Uses


### Direct Use

Load the model, finetune the model if needed or just go straight towards generating inferences using the model.



### Downstream Use

Finetune the model on other diagnostic scans, though the model only accepts grayscale images of size 256x256.

## How to Get Started with the Model

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1SfK9d2In3JHDvyXH4jpwznVGEG_wXRuQ?usp=sharing)

## Training
The colab notebook used to train the model can be found below

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1SfK9d2In3JHDvyXH4jpwznVGEG_wXRuQ?usp=sharing)

## Evaluation

### Metrics

| Class       | Precision | Recall | F1-Score | Support |
|-------------|-----------|--------|----------|---------|
| Glioma      | 0.96      | 0.87   | 0.91     | 300     |
| Meningioma  | 0.84      | 0.71   | 0.77     | 306     |
| No Tumor    | 0.88      | 1.00   | 0.93     | 405     |
| Pituitary   | 0.93      | 0.99   | 0.96     | 300     |
| **Accuracy**|           |        | **0.90** | 1311    |
| **Macro Avg** | 0.90    | 0.89   | 0.89     | 1311    |
| **Weighted Avg** | 0.90 | 0.90   | 0.90     | 1311    |

![Roc curve](https://raw.githubusercontent.com/Kaynaaf/BrainMRI-Classifier/main/results/roc_curve.png)
### Results
This model was developed for my project that can be found on github [here](https://github.com/Kaynaaf/BrainMRI-Classifier)
. This project involved generating sensitivity maps to explain the predictions of the model.
These maps assign values to areas of the image that act as feature importance markers.
![pituitary Tumour sensitivity map](https://raw.githubusercontent.com/Kaynaaf/BrainMRI-Classifier/main/results/pituitary_1.png)
![Glioma  sensitivity map](https://raw.githubusercontent.com/Kaynaaf/BrainMRI-Classifier/main/results/glioma_1.png)