High‐speed Intraoperative Assessment of Breast Tumor Margins by Multimodal Ultrasound and Photoacoustic Tomography

Abstract

Conventional methods for breast tumor margins assessment need a long turnaround time, which may lead to re‐operation for patients undergoing lumpectomy surgeries. Photoacoustic tomography (PAT) has been shown to visualize adipose tissue in small animals and human breast. Here, we demonstrate a customized multimodal ultrasound and PAT system for intraoperative breast tumor margins assessment using fresh lumpectomy specimens from 66 patients. The system provides the margin status of the entire excised tissue within 10 minutes. By subjective reading of three researchers, the results show 85.7% [95% confidence interval (CI), 42.0% ‐ 99.2%] sensitivity and 84.6% (95% CI, 53.7% ‐ 97.3%) specificity, 71.4% (95% CI, 30.3% ‐ 94.9%) sensitivity and 92.3% (95% CI, 62.1% ‐ 99.6%) specificity, and 100% (95% CI, 56.1% ‐ 100%) sensitivity and 53.9% (95% CI, 26.1% ‐ 79.6%) specificity respectively when cross‐correlated with post‐operational histology. Furthermore, a machine learning‐based algorithm is deployed for margin assessment in the challenging ductal carcinoma in situ tissues, and achieved 85.5% (95% CI, 75.2% ‐ 92.2%) sensitivity and 90% (95% CI, 79.9% ‐ 95.5%) specificity. Such results present the potential of using mutlimodal ultrasound and PAT as a high‐speed and accurate method for intraoperative breast tumor margins evaluation.