A proof-of-concept study of optical tissue sensing during vacuum-assisted breast biopsy procedures.

Abstract

Background: Vacuum-assisted breast biopsy (VABB) enables minimally invasive sampling of breast lesions but provides no real-time information about the tissue being sampled, potentially leading to sampling bias. We evaluated whether integrating diffuse reflectance spectroscopy (DRS) into the VABB workflow is feasible and can support more targeted sampling.

Methods: A fiber-equipped introducer compatible with a standard Hologic VABB system was developed to acquire DRS measurements without altering the clinical workflow. In 40 ex vivo mastectomy specimens, ultrasound-guided VABB was performed, with DRS measurements obtained immediately before biopsy acquisition. Histopathology served as the reference standard, and labeled spectra were used to train machine learning models using sample-wise and patient-wise classification to distinguish healthy from malignant tissue.

Results: A total of 300 biopsies were correlated with histopathology, yielding a balanced dataset of 248 spectra. The sample-wise model achieved a mean accuracy of 86%, sensitivity of 91%, and specificity of 82%. The patient-wise approach achieved 76%, 85%, and 70%, respectively.

Conclusions: This proof-of-concept study demonstrates that DRS can be integrated into the VABB workflow using a novel optical introducer and discriminate malignant from healthy tissue in ex vivo breast specimens, supporting technical feasibility. Further clinical studies are needed to establish its diagnostic value.