Detection of breast cancer using machine learning on time-series diffuse optical transillumination data

Area of Science:

• Biomedical optics
• Machine learning applications
• Medical imaging analysis

Background:

• Optical mammography has not met expectations for cancer diagnosis.
• Machine learning (ML) presents an opportunity to enhance cancer detection in diffuse optical transmission data.

Purpose of the Study:

• To quantitatively assess ML methods for classifying cancer-positive versus cancer-negative patients.
• To evaluate ML performance on raw transmission time series data from bilateral breast scans.

Main Methods:

• Utilized a support vector machine (SVM) with hyperparameter optimization and cross-validation.
• Employed an automated ML (AutoML) framework for validation.
• Quantified classification performance using receiver operating characteristics and area under the curve (AUC).

Main Results:

• Achieved an AUC score of up to 93.3% for SVM classification.
• Attained an AUC score of up to 95.0% for the AutoML classifier.
• Demonstrated high diagnostic performance on a sample group of 18 cancer patients.

Conclusions:

• ML provides a viable strategy for clinically relevant breast cancer diagnosis via diffuse optical transmission.
• ML on raw data can surpass traditional statistical biomarkers from reconstructed images.
• Simultaneous bilateral scanning with dense channel coverage is crucial for clinical relevance.