APPLICATION OF DEEP LEARNING TECHNIQUES FOR BREAST CANCER CELLS DETECTION USING A FIBER-OPTIC BIOSENSOR

Abstract

Biosensor-based detection is a new field of research that has great potential in early-stage breast cancer detection and treatment. This novel approach can help solve the issue of delayed cancer diagnoses, reduce false positives and establish a safe, fast, and cost-effective diagnosis method. This work focuses on an in-house fabricated biosensor and discusses its results from conducted analyte detection experiments, including CD44-expressing cancer cells. The experiments aimed to explore the sensitivity, specificity, and Limits of Detection (LoD) of the fabricated sensors in response to cells with different levels of CD44, a glycoprotein present in many forms of cancerous cells. Those metrics help identify the repeatability and stability of the sensors together with the lowest limit of cell concentration that a sensor can detect. This study aimed to investigate ways to advance the analysis of cancer cell detection experiments by using Generative Artificial Intelligence for data augmentation and Machine Learning for cell concentration classification. The results achieved during this work included an increase in the classification accuracy for cell concentration in media from 36% to 77% using Random Forest model. Further improvement and investigation of this work will help pave the way for improved reliability and preciseness in detection.