Rapid identification of breast cancer subtypes using micro-FTIR and machine learning methods

Area of Science:

• Biomedical optics
• Computational biology
• Cancer research

Background:

• Fourier transform infrared (FTIR) spectroscopic imaging offers label-free biochemical analysis for breast cancer (BC) diagnosis.
• Current FTIR methods for BC subtype diagnosis are hindered by long acquisition times and computational challenges, limiting clinical utility.
• Machine learning (ML) presents a promising avenue for accelerating BC subtype classification and improving diagnostic accuracy.

Purpose of the Study:

• To develop and validate a novel ML-based computational framework for rapid and accurate classification of breast cancer molecular subtypes.
• To enhance the clinical feasibility of FTIR spectroscopic imaging for BC diagnosis by addressing limitations in data acquisition speed and computational processing.
• To improve the accuracy, specificity, and sensitivity of BC subtype identification using an integrated FTIR imaging and ML approach.

Main Methods:

• Development of a novel machine learning algorithm, Neighborhood Components Analysis-K-Nearest Neighbors (NCA-KNN), for computational distinction of BC cell lines.
• Integration of FTIR spectroscopic imaging data with the NCA-KNN algorithm to enable label-free biochemical analysis.
• Comparative analysis of the proposed NCA-KNN method against a supervised support vector machine (SVM) model for BC subtype classification.

Main Results:

• The NCA-KNN method achieved high classification accuracy (97.5%), specificity (96.3%), and sensitivity (98.2%) for BC subtypes using FTIR imaging data.
• Significant improvements in diagnostic performance were observed even with very low co-added scans and short acquisition times.
• The proposed NCA-KNN method demonstrated a distinct accuracy advantage (up to 9%) over the best-performing SVM model.

Conclusions:

• The developed NCA-KNN method offers a computationally efficient and highly accurate approach for breast cancer subtype classification.
• This ML-driven FTIR imaging technique shows potential for overcoming current clinical adoption barriers, enabling faster and more precise BC diagnosis.
• The NCA-KNN method represents a significant advancement for subtype-associated therapeutics and personalized medicine in breast cancer treatment.