https://github.com/msikorski93/pneumonia-recognition-with-deep-learning-and-tensorflow

A binary classification using Convolution Neural Network (CNN, or ConvNet) model.
https://github.com/msikorski93/pneumonia-recognition-with-deep-learning-and-tensorflow

convolutional-neural-network efficientnet grad-cam healthcare-analysis image-classification inceptionv3 keras pneumonia-detection resnet-50 tensorflow transfer-learning vgg16 vgg19

Last synced: 3 days ago
JSON representation

A binary classification using Convolution Neural Network (CNN, or ConvNet) model.

• Host: GitHub
• URL: https://github.com/msikorski93/pneumonia-recognition-with-deep-learning-and-tensorflow
• Owner: msikorski93
• Created: 2023-01-23T13:53:32.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2024-07-25T12:49:35.000Z (4 months ago)
• Last Synced: 2024-07-25T14:28:04.488Z (4 months ago)
• Topics: convolutional-neural-network, efficientnet, grad-cam, healthcare-analysis, image-classification, inceptionv3, keras, pneumonia-detection, resnet-50, tensorflow, transfer-learning, vgg16, vgg19
• Language: Jupyter Notebook
• Homepage:
• Size: 6.05 MB
• Stars: 3
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Pneumonia-Recognition-With-Deep-Learning-and-TensorFlow

![ alt text ](https://img.shields.io/badge/license-MIT-green?style=&logo=)

![ alt text ](https://img.shields.io/badge/-Jupyter-F37626?logo=Jupyter&logoColor=white)

![ alt text ](https://img.shields.io/badge/-NumPy-013243?logo=Numpy&logoColor=white)

![ alt text ](https://img.shields.io/badge/-TensorFlow-FF6F00?logo=TensorFlow&logoColor=white)

![ alt text ](https://img.shields.io/badge/-Keras-D00000?logo=Keras&logoColor=white)

![ alt text ](https://img.shields.io/badge/-scikit--learn-F7931E?logo=scikitlearn&logoColor=white)

In this project a convolutional neural network (CNN) was built and trained in Keras with TensorFlow as backend from scratch to predict

if patients were infected with pneumonia using their chest radiology images. The dataset contained the lungs X-ray images of both groups: normal

and pneumonia infected patients. The model used was sequential with a combination of convolutional layers, pooling layers, dropout layers, dense layers

with ReLu activation and output layer with sigmoid activation. The CNN was designed with the following architecture:

![architecture](https://user-images.githubusercontent.com/45270023/214060460-810941ae-b8e6-43c9-9a9e-695520737e33.jpg)

The learning process stopped at the 33rd epoch. The model evaluation on test data achieved 86.06% for accuracy and 93.34% for the AUC score. Finally, we

made a single image recognition on an online chest image. The algorithm performs correctly in terms of classifying patients radiology images.

### New notebook for November 2023

A new classification task was performed using transfer learning. Five different pre-trained models were generated and performed with the following evaluation metrics:

|                | accuracy | recall   | precision | auc      | f1_score | f2_score |

|----------------|----------|----------|-----------|----------|----------|----------|

| ResNet50       | 0.832532 | 0.830128 | 0.834139  | 0.875502 | 0.967742 | 0.949367 |

| VGG16          | 0.883814 | 0.889423 | 0.879556  | 0.938604 | 0.875000 | 0.875000 |

| VGG19          | 0.818109 | 0.823718 | 0.814580  | 0.904382 | 0.848485 | 0.864197 |

| InceptionV3    | 0.828526 | 0.828526 | 0.828526  | 0.853080 | 0.687500 | 0.687500 |

| EfficientNetB1 | 0.783654 | 0.783654 | 0.783654  | 0.882095 | 0.875000 | 0.875000 |

This notebook has been additionally developed with Grad-CAM visualizations for the best CNN - VGG16.