Body Part Recognition System

This project implements a body part recognition system using machine learning techniques. It processes DICOM images, converts them to PNG format, and utilizes SVM (Support Vector Machine) and KNN (K-Nearest Neighbors) classifiers to identify various body parts from medical images.

This project is based on my other project TrafficSignRecognition.

Features

• Model Training: Train classification models using SVM and KNN algorithms.
• Prediction: Make predictions on test datasets and evaluate model performance.
• Data Handling: Load training and testing data from specified directories.
• DICOM Image Processing: Load and preprocess DICOM images for analysis.

Getting Started

Prerequisites

• C++ Compiler (e.g., g++)
• CMake
• OpenCV
• DCMTK (for DICOM file handling)

Demo

1. Loading data

2. SVM Model

3. KNN Model

Installation

1. Clone the repository:

   • git clone https://github.com/amWRit/BodyPartRecognition.git
   • cd BodyPartRecognition

2. Install the required dependencies (OpenCV, DCMTK).

Usage

1. Prepare your dataset:

• Place your training data in the bodyparts-train-data directory.
• Place your test png images in the images directory.
• Place your test dicom images in the dicom-images directory.

2. Run the application: ./BodyPartRecognition

3. The application will load the data, train the models, and output predictions along with accuracy metrics.

Models Used

• SVM (Support Vector Machine): A supervised learning model used for classification tasks.
• KNN (K-Nearest Neighbors): A simple, instance-based learning algorithm used for classification.

Acknowledgments

• X-ray Body Images Dataset
• DCMTK Documentation

License

This project is licensed under the MIT License - see the LICENSE file for details.