Knee Osteoarthritis Detection and KL Grading using AlexNet5 and Xception Models

This project is aimed at developing deep learning models for the detection of knee osteoarthritis and grading the Kellgren-Lawrence (KL) severity using two models: a manually trained AlexNet5 and a fine-tuned Xception model.

Dataset

The project uses a dataset of digital knee X-ray images, which was obtained from Mendeley Data with the following citation:

Gornale, Shivanand; Patravali, Pooja (2020), “Digital Knee X-ray Images”, Mendeley Data, V1, doi: 10.17632/t9ndx37v5h.1

The dataset consists of 1650 digital X-ray images of knee joint which are collected from well reputed hospitals and diagnostic centres. The X-ray images are acquired using PROTEC PRS 500E X-ray machine. Original images are 8-bit grayscale image. Each radiographic knee X-ray image is manually annotated /labelled as per Kellgren and Lawrence grades(0-4).

Prerequisites

• Python 3.10.0
• TensorFlow 2.10.1
• Bazel 5.1.1
• cuDNN 8.1
• CUDA 11.2
• Keras
• Numpy
• Pandas
• Scikit-learn
• Matplotlib
• Seaborn
• openCV

Installation

1. Clone the repository to your local machine.
2. Install the required packages using the following command:

pip install -r requirements.txt

Usage

Model Training and Preprocessing

1. Open the 'model.ipynb' notebook using Jupyter Notebook or JupyterLab.
2. Follow the instructions in the notebook to preprocess the data and train the models.

Testing

1. Open the 'XrayChecker.ipynb' notebook using Jupyter Notebook or JupyterLab.
2. Follow the instructions in the notebook to test the trained models on new data.

Model Saving and Logs

1. After training the models, they will be saved in the current directory as 'AlexNet5.hdf5' and 'XceptionModel.hdf5'.
2. The training logs for both models will be saved in the current directory as alexnet5_history.log and xception_history.log.

Results

1. The manually trained AlexNet5 model achieved a training accuracy of 43.59% and a test accuracy of 40.00% for multi-class classification of KL classification 0-4.

2. The fine-tuned Xception model achieved a training accuracy of 83.32% and a test accuracy of 82.67% for the same multi-class KL classification.

Website

Landing Page	About

Application	Application Working

Credits

This project was developed by [Subhrajit Panda].

License

This project is licensed under the Apache License, Version 2.0 - see the LICENSE file for details.