Performance Indicators for Shape and Position Assessment in Electromagnetic Inverse Scattering

Overview

This repository contains the complete research materials for the paper "Performance Indicators for Shape and Position Assessment in Electromagnetic Inverse Scattering" submitted to IEEE Transactions on Antennas and Propagation.

The research introduces two novel performance indicators specifically designed to evaluate electromagnetic inverse scattering algorithms:

• Shape Error Indicator: Quantifies geometric reconstruction accuracy
• Position Error Indicator: Measures object localization precision

Abstract

Microwave imaging is a low-cost, non-invasive technique for detecting and characterizing objects in inaccessible media through electromagnetic field measurements. The evaluation of algorithms that solve the related electromagnetic inverse scattering problem typically relies on metrics such as Mean Square Error (MSE) or Structural Similarity (SSIM), which primarily assess contrast estimation accuracy but provide limited insight into the recovery of object geometry and position.

This paper introduces two novel performance indicators specifically designed to evaluate electromagnetic inverse scattering algorithms. The proposed indicators are applicable to both qualitative and quantitative methods, even when they are combined in an experiment. Comprehensive experimental validation is conducted using traditional algorithms: Linear Sampling Method (LSM), Orthogonality Sampling Method (OSM), Born Iterative Method (BIM), Contrast Source Iterative Method (CSI), Subspace Optimization Method (SOM), and Circle Approximation (CA).

Repository Structure

PISPAEIS/
├── experiments/                    # Experimental implementations and results
│   ├── breast/                     # Breast phantom experiments
│   ├── shape/                      # Shape recovery studies
│   └── position/                   # Position detection studies
├── eispy2d/                        # Core algorithm implementations
│   └── library/                    # EISPY2D electromagnetic inverse scattering library
├── data/                           # Experimental datasets
│   ├── breast/                     # Breast phantom data
│   ├── shape/                      # Shape experiment data
│   └── position/                   # Position experiment data
├── requirements.txt                # Python dependencies
├── loaddata.py                     # Data loading utilities
└── README.md                       # This file

Key Contributions

1. Novel Performance Indicators: Introduction of shape and position error indicators for electromagnetic inverse scattering evaluation
2. Comprehensive Algorithm Comparison: Evaluation of six traditional algorithms (LSM, OSM, BIM, CSI, SOM, CA)
3. Multiple Experimental Frameworks: Three distinct experimental setups demonstrating practical applicability
4. Statistical Validation: Robust statistical analysis supporting the experimental findings
5. Open-Source Implementation: Complete code and data for reproducibility

Algorithms Evaluated

• LSM: Linear Sampling Method
• OSM: Orthogonality Sampling Method
• BIM: Born Iterative Method
• CSI: Contrast Source Iterative Method
• SOM: Subspace Optimization Method
• CA: Circle Approximation

Key Findings

The experiments revealed that:

• Shape Recovery: In scenarios with a single scatterer and low Degree of Nonlinearity, BIM achieves the best average performance in object geometry recovery
• Position Detection: In scenarios with a single small-sized, high-contrast scatterer, SOM and OSM achieve the best average performance in target localization

Installation and Setup

Prerequisites

• Python 3.12+
• LaTeX distribution (for paper compilation)
• Required Python packages (see requirements.txt)

Environment Setup

1. Clone the repository:

   git clone https://github.com/andre-batista/PISPAEIS.git
   cd PISPAEIS

2. Create virtual environment:

   python -m venv env
   source env/bin/activate  # On Windows: env\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

Running Experiments

1. Load experimental data:

   python loaddata.py

2. Run specific experiments:

   # Shape recovery experiments
   cd experiments/shape/star
   python runexperiment.py
   
   # Position detection experiments
   cd experiments/position/single
   python runexperiment.py
   
   # Breast phantom experiments
   cd experiments/breast
   python runexperiment.py

Data Description

Experimental Datasets

• Shape Recovery Data: Various geometric shapes with different complexity levels
• Position Detection Data: Objects at different spatial locations
• Breast Phantom Data: Realistic medical imaging scenarios

Data Format

All experimental data is stored in standardized formats compatible with the EISPY2D library.

Citation

If you use this work in your research, please cite:

@article{batista2025performance,
    title={Performance Indicators for Shape and Position Assessment in Electromagnetic Inverse Scattering},
    journal={IEEE Access}, 
    author={Batista, André Costa and Adriano, Ricardo},
    year={2026},
}

Authors

• André Costa Batista - Department of Electrical Engineering, Universidade Federal de Minas Gerais
• Ricardo Adriano - Department of Electrical Engineering, Universidade Federal de Minas Gerais

Acknowledgments

This work was supported in part by:

• Brazilian agency CAPES (Coordination for the Improvement of Higher Education Personnel) under Grant 88887.463864/2019-00
• FAPEMIG-CNPQ scholarship (process APQ-06716-24)
• CNPq (The National Council for Scientific and Technological Development)

We also thank Lucas S. Batista for his guidance and support during the research.

License

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

Contact

For questions about this research, please contact:

• André Costa Batista: andre-costa@ufmg.br

---

Keywords: Algorithm evaluation, electromagnetic inverse scattering, microwave imaging, performance indicators