A core needle biopsy combined with novel spectroscopic probe for in vivo tissue classification – A pilot study on piglets

Abstract

Tissue sampling is a primary goal of core needle biopsies (CNB), cancer therapy evaluation, and autoimmune disease assessment. Conventional guidance methods such as ultrasound and MRI suffer from periprocedural tissue‐type insensitivity in complex biopsy targets, motion sensitivity, imaging artifacts and high costs, which may limit their usefulness. Accurate tissue classification and needle guidance during CNB are equally important. Mistakes may lead to sample inadequacies, obscured results, incorrect sampling spots, and ultimately repeated biopsies. To address these challenges, this study investigates the feasibility of a smart CNB probe integrating real-time optical spectroscopy for enhanced tissue characterization during in vivo biopsy utilizing machine learning methods. Ten fabricated probes were tested in vivo on porcine fat, liver, and kidney tissues, demonstrating potential for improving biopsy accuracy. Acquired spectral data enabled effective tissue differentiation, as indicated by the best-performing classification models. LDA classifier with MRMR feature selection reached sensitivity of 87.3 % in classification between liver and fat tissues, where SVM with linear kernel and PCA reached 86.4 % sensitivity in kidney vs fat. These findings suggest that integrating optical spectroscopy into CNB procedures may enhance diagnostic accuracy while mitigating procedural risks.