Carbon Nanoparticles in Breast Cell Imaging for CNN-Based Tumor Detection

Michela Gravina, Salvatore Capuozzo, Gaetano Saviano, Julian Gortz, Valeria Panzetta, Carmela Russo, Mariano Sirignano, Paolo Antonio Netti, Carlo Sansone

Abstract: In recent years, the intersection of Artificial Intelligence (AI) and fluorescence microscopy has reshaped biomedical imaging, enabling the scalable, automated, and high-precision analysis of complex cellular structures. In parallel, Carbon Dots (CDs) have emerged as promising nanomaterials for both imaging enhancement and modulation of cellular properties. In this study, we investigate the role of CDs in AI-based analysis, proposing a modular deep learning pipeline for the classification of healthy and cancerous breast cells from fluorescence microscopy images. The framework integrates unsupervised cell segmentation with a Convolutional Neural Network (CNN)-based classification model, enabling a systematic analysis of cells in their native form (Control scenario) and after exposure to CDs (Exposed scenario). Experimental results, supported by explainable AI techniques, show that CDs not only enhance image contrast but also facilitate the learning of more discriminative and transferable features, improving classification performance across both intra-scenario and cross-scenario evaluations.

External IDs:doi:10.1007/978-3-032-11381-8_6