# Advanced image reconstruction algorithms for high-resolution digital time-of-flight PET/CT enhance visualization of sub-clinical internal mammary lymph node metastases in breast cancer: a phantom and a clinical, retrospective cohort study

## TL;DR

New PET/CT image reconstruction methods improve detection of small breast cancer lymph node metastases.

## Contribution

Advanced HYPER Iterative and uAI HYPER DPR reconstruction algorithms enhance visualization of internal mammary lymph node metastases in breast cancer.

## Key findings

- HYPER Iterative and uAI HYPER DPR improve lesion conspicuity and contrast in IMLN metastases.
- Higher SUVmax and tumor-to-background ratios were observed with these advanced reconstruction methods.
- Diagnostic confidence was higher with HYPER Iterative and uAI HYPER DPR compared to OSEM.

## Abstract

Internal mammary lymph node (IMLN) metastases play an important role in breast cancer staging and treatment planning but is often difficult to detect because of their small size and anatomical location. Recent advances in digital time-of-flight (TOF) positron emission tomography (PET)/CT and advanced image reconstruction techniques may improve the visualization of such small lesions. This study aimed to evaluate the performance of advanced reconstruction methods (HYPER Iterative and uAI HYPER DPR) for visualizing IMLN metastases in breast cancer using phantom and clinical data.

A modified NEMA image quality phantom and a retrospective cohort of breast cancer patients with IMLN metastases were evaluated using a high-resolution digital TOF PET/CT system (uMI 550). Images were reconstructed using ordered subset expectation maximization (OSEM), HYPER Iterative, and uAI HYPER DPR with different reconstruction parameters, and quantitative metrics and visual scores were assessed.

In both phantom and clinical images, smaller RS-values for HYPER Iterative and larger Str-values for uAI HYPER DPR were associated with higher lesion conspicuity and contrast-related metrics, at the expense of increased noise. Images reconstructed with a 256 × 256 matrix showed lower background variability than those reconstructed with a 512 × 512 matrix. In the clinical study, these reconstruction settings resulted in higher SUVmax and tumor-to-background ratios for IMLN metastases, and visual scores for diagnostic confidence were higher for HYPER Iterative (RS = 0.7–0.91) and uAI HYPER DPR (Str = 2–4) than for OSEM.

The online version contains supplementary material available at 10.1186/s40658-026-00846-8.