SIR-3DCNN: A Framework of Multivariate Time Series Classification for Lung Cancer Detection

Overview

This repository contains the code and resources for the paper "SIR-3DCNN: A Framework of Multivariate Time Series Classification for Lung Cancer Detection". The project introduces a novel framework that combines Sensor Array Optimization, Spatiotemporal Information Representation, and a 3D Convolutional Neural Network (3DCNN) to classify multivariate time series data for early detection of lung cancer.

Table of Contents

• Project Structure
• Installation
• Usage
  • Sensor Array Optimization
  • Spatiotemporal Information Representation
  • Classification
• Reference

Project Structure

• classification/: Contains scripts for building and training the 3DCNN model.
  • models.py: Defines the architecture of the 3DCNN.
  • train.py: Training script for the model.
• media/: Contains images and diagrams.
  • framework.png: Visual representation of the framework.
• sensor_array_optimization/: Includes scripts and results for sensor selection.
  • SAO_LDA.py: Performs sensor array optimization using LDA.
  • array_optimization_result.txt: Results of the sensor optimization.
• spatiotemporal_information_representation/: Scripts for data transformation.
  • SIA.py: Obtains the optimal spatiotemporal representation.
• metrics.py: Contains functions for evaluating model performance.
• README.md: Project documentation.

Installation

Prerequisites

• Python 3.7 or higher
• Required Python packages listed in requirements.txt

Steps

1. Clone the repository

   git clone https://github.com/cqu-3dteam/sir-3dcnn.git

2. Install dependencies

   cd sir-3dcnn
   
   pip install -r requirements.txt
   

Usage

Sensor Array Optimization

python sensor_array_optimization/SAO_LDA.py

Spatiotemporal Information Representation

python spatiotemporal_information_representation/SIA.py

Classification

python classification/train.py

Reference

@ARTICLE{SIR-3DCNN,
  author={Ran Liu, Shidan Wang, Fengchun Tian, Lin Yi},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  title={SIR-3DCNN: A Framework of Multivariate Time Series Classification for Lung Cancer Detection}, 
  year={2024},
  volume={},
  number={},
  pages={},
  doi={}
}