# Innovative Discrete Multi-Wavelength Near-Infrared Spectroscopic (DMW-NIRS) Imaging for Rapid Breast Lesion Differentiation: Feasibility Study

## TL;DR

This study shows that a new near-infrared imaging device can help distinguish between benign and malignant breast lesions more accurately than traditional methods.

## Contribution

The study introduces a novel DMW-NIRS imaging device that improves breast lesion differentiation when combined with ultrasound.

## Key findings

- TOIL/N ratios achieved the highest diagnostic accuracy for differentiating malignant from benign lesions.
- Adaptive BI-RADS models using DMW-NIRS outperformed initial radiologist assessments for BI-RADS 4A lesions.
- The device offers a non-invasive, rapid method to reduce unnecessary breast biopsies.

## Abstract

Background/Objectives: This study evaluated the role of a discrete multi-wavelength near-infrared spectroscopic (DMW-NIRS) imaging device for rapid breast lesion differentiation. Methods: A total of 62 women (mean age, 49.9 years) with ultrasound (US)-guided biopsy-confirmed breast lesions (37 malignant, 25 benign) were included. A handheld probe equipped with five pairs of light-emitting diodes (LEDs) and photodiodes (PDs) measured lesion-to-normal tissue (L/N) ratios of four chromophores, THC (Total Hemoglobin Concentration), StO2, and the Tissue Optical Index (TOI: log10(THC × Water/Lipid)). Lesions were localized using US. Diagnostic performance was assessed for each L/N ratio, with subgroup analysis for BI-RADS 4A lesions. Two adaptive BI-RADS models were developed: Model 1 used TOIL/N thresholds (Youden index), while Model 2 incorporated radiologists’ reassessments of US findings integrated with DMW-NIRS results. These models were compared to the initial BI-RADS assessments, conducted by breast-dedicated radiologists. Results: All L/N ratios significantly differentiated malignant from benign lesions (p < 0.05), with TOIL/N achieving the highest AUC-ROC (0.901; 95% CI: 0.825–0.976). In BI-RADS 4A lesions, all L/N ratios except Lipid significantly differentiated malignancy (p < 0.05), with TOIL/N achieving the highest AUC-ROC (0.902; 95% CI: 0.788–1.000). Model 1 and Model 2 showed superior diagnostic performance (AUC-ROCs: 0.962 and 0.922, respectively), significantly outperforming initial BI-RADS assessments (prospective AUC-ROC: 0.862; retrospective AUC-ROC: 0.866; p < 0.05). Conclusions: Integrating DMW-NIRS findings with US evaluations enhances diagnostic accuracy, particularly for BI-RADS 4A lesions. This novel device offers a rapid, non-invasive, and efficient method to reduce unnecessary biopsies and improve breast cancer diagnostics. Further validation in larger cohorts is warranted.