Breast cancer differential diagnosis using diffuse optical spectroscopic imaging and regression with z-score...

Area of Science:

• Biomedical optics
• Medical imaging
• Breast cancer diagnostics

Background:

• Current breast lesion diagnosis has high false positive rates, leading to unnecessary biopsies.
• Diffuse optical spectroscopic imaging (DOSI) offers noninvasive assessment of tissue hemodynamics and composition.
• DOSI is being explored to improve accuracy, especially in dense breasts.

Purpose of the Study:

• To evaluate the efficacy of DOSI in differentiating malignant breast lesions from benign ones.
• To assess the impact of z-score normalization on DOSI model performance.
• To determine if DOSI can serve as an adjunct to reduce invasive biopsies.

Main Methods:

• Retrospective analysis of 212 subjects with suspicious breast lesions undergoing DOSI.
• Z-score normalization of physiological tissue parameters against contralateral breast tissue.
• Univariate logistic regression models to discriminate between malignant and benign lesions.

Main Results:

• Models with hemodynamic parameters achieved an area under the curve (AUC) ≥0.85 for distinguishing malignant tumors.
• Z-score normalization enhanced model discriminatory ability and calibration compared to other normalization methods.
• Accurate differentiation of malignant from surrounding tissue and benign lesions was achieved.

Conclusions:

• Combining DOSI data normalization with statistical regression improves diagnostic accuracy for malignant lesions.
• DOSI's portability, low cost, and nonionizing nature make it a practical adjunct tool.
• Improved DOSI accuracy can significantly reduce the need for unnecessary invasive breast biopsies.