Evaluating Detection Thresholds: The Impact of False Positives and Negatives on Super-Resolution Ultrasound Localization Microscopy

TL;DR

The paper addresses how detection thresholds for microbubble identification in Ultrasound Localization Microscopy impact SR map quality. It uses controlled simulations with two center frequencies ($f_1=2.841$ MHz and $f_2=7.24$ MHz) to inject false positives ($FP$) and false negatives ($FN$), evaluating SR maps via $SSIM$ and $PSNR$. A density-based region analysis, using Gaussian KDE to separate dense and sparse MB regions, reveals that $FN$ errors cause substantially larger degradations in $SSIM$ than $FP$ errors, especially in sparse regions and at higher frequencies. The findings advocate for adaptive, region-aware detection thresholds to balance accuracy and robustness, with particular attention to sparse regions and higher-frequency imaging to preserve structural fidelity in SR maps.

Abstract

Super-resolution ultrasound imaging with ultrasound localization microscopy (ULM) offers a high-resolution view of microvascular structures. Yet, ULM image quality heavily relies on precise microbubble (MB) detection. Despite the crucial role of localization algorithms, there has been limited focus on the practical pitfalls in MB detection tasks such as setting the detection threshold. This study examines how False Positives (FPs) and False Negatives (FNs) affect ULM image quality by systematically adding controlled detection errors to simulated data. Results indicate that while both FP and FN rates impact Peak Signal-to-Noise Ratio (PSNR) similarly, increasing FP rates from 0\% to 20\% decreases Structural Similarity Index (SSIM) by 7\%, whereas same FN rates cause a greater drop of around 45\%. Moreover, dense MB regions are more resilient to detection errors, while sparse regions show high sensitivity, showcasing the need for robust MB detection frameworks to enhance super-resolution imaging.

Figures (3)

• Figure 1: left: SR Maps, middle: KDE analysis and right: Region Masks (black: Sparse and white: Dense regions) for (A) Simulation 1 and (B) Simulation 2
• Figure 2: Result of varying FP and FNs in Sparse, dense and all the regions of image (indicated by columns) for (A) SSIM (B) PSNR of Simulation 1
• Figure 3: Result of varying FP and FNs in Sparse, dense and all the regions of image (indicated by columns) for (A) SSIM and (B) PSNR of Simulation 2.