Small breast lesion classification performance using the normalized axial-shear strain area feature.

Area of Science:

• Medical imaging
• Ultrasound technology
• Oncology

Background:

• Symptomatic breast cancers are often larger and have metastasized.
• Early detection and confirmation of breast lesions are critical for effective treatment.
• Normalized axial-shear strain area (NASSA) is a promising feature for identifying cancerous boundary-bonding conditions.

Purpose of the Study:

• To investigate the size distribution of breast lesions in a previous study.
• To analyze the classification performance of the NASSA feature on small lesions (<10 mm).
• To compare the classification accuracy of NASSA for small versus large lesions.

Main Methods:

• Analysis of 33 malignant tumors and 30 fibroadenomas.
• Observers outlined lesions; size was computed using maximum circle-equivalent diameter.
• Axial-shear strain elastography (ASSE) was segmented, and NASSA computed semi-automatically.
• Receiver operating characteristic (ROC) curves generated for small lesions; box plots for size groups.

Main Results:

• Approximately 38% of fibroadenomas and 22% of cancers were small (<10 mm).
• NASSA achieved perfect classification of small lesions in training and cross-validation.
• Small lesions (<10 mm) showed better classification performance than larger lesions (>10 mm).
• Mean score differences between fibroadenomas and cancers were significant for both small and large lesions.

Conclusions:

• The ASSE-derived NASSA feature demonstrates high accuracy in classifying small breast lesions.
• NASSA performs effectively on small lesions, comparable to or better than larger ones.
• This feature can enhance standard ultrasound Breast Imaging Reporting and Data System (BIRADS) classification for improved breast cancer diagnosis.