This repository contains code and scripts used for analyzing the effect of detection thresholds—specifically false positives and false negatives—on Super-Resolution Ultrasound (SR-US) and Ultrasound Localization Microscopy (ULM) image quality.
The focus of this work is to quantitatively study how microbubble detection errors propagate into super-resolution density maps and affect commonly used image quality metrics.
The datasets used are from UltraSR challenge and can be downloaded from Zenodo:
🔗 https://zenodo.org/records/7271766
Please download the data separately and place it in the appropriate directory structure as expected by the scripts in this repository.
- Scripts for modifying microbubble detection outputs by:
- Adding controlled numbers of false positives
- Introducing false negatives at specified rates
- Generation of super-resolution density maps
- Quantitative evaluation using image quality metrics such as:
- Structural Similarity Index (SSIM)
- Dice coefficient
- Peak Signal-to-Noise Ratio (PSNR)
- Analysis of dense vs. sparse regions in SR maps
The code is intended for research and reproducibility purposes and may require adaptation for different datasets or experimental setups.
If you use this code or data in your work, please cite the following paper for now:
ArXiv preprint
https://arxiv.org/abs/2411.07426
This work has been accepted to SPIE Medical Imaging 2026.
The citation information will be updated once the final published version becomes available.
For questions, issues, or collaboration inquiries, please open an issue on GitHub or contact the authors directly.
This repository is intended for academic and research use.
Please check the Zenodo record for data-specific licensing information.