Can artificial intelligence and contrast-enhanced mammography be of value in the assessment and characterization of breast lesions?

Abstract

Abstract Background Breast imaging plays a crucial role in the early detection of breast cancer, greatly contributing to improved management, higher cure rates and a significant reduction in mortality. Mammography is the gold standard for breast screening yet, and it has low sensitivity and specificity in the identification and diagnosis of breast lesions, particularly in dense breasts. Radiologists, under heavy and prolonged workloads, are more prone to errors and to reduce such mistakes, and computer-aided diagnosis (CAD) has been introduced. Artificial intelligence (AI) can be used as a second reader to mammographic images, thus decreasing recalls while improving cancer detection rates. On the other hand, contrast-enhanced mammography (CEM) is an imaging technique that provides enhanced morphological information in addition to functional data. We aimed to evaluate the accuracy of AI software algorithms in the assessment and characterization of breast lesions compared to CEM. Results This prospective study included 58 patients with a total of 74 lesions who underwent either screening or diagnostic digital mammography. Each participant underwent full-field digital mammography, ultrasound and CEM. The resulting mammographic images were processed using AI algorithm. In our study, CEM demonstrated a sensitivity of 98.33%, specificity of 92.86%, positive predictive value (PPV) of 98.34%, negative predictive value (NPP) of 92.85%, and accuracy of 97.3%. In comparison, AI showed a sensitivity of 91.67%, specificity of 85.71%, PPV of 96.5%, NPP of 70.56%, and accuracy of 90.54%. In addition, CEM detected multifocality in 100% and multicentricity in 100% of cases compared to 11.11% and 50% in AI, respectively. Conclusions The overall diagnostic indices of AI on digital mammogram were comparable to mammography and ultrasonography results. As such, it could serve as a valuable optional complement for assessing and characterizing breast lesions and act as a second reader for digital mammographic images. On the other hand, CEM will still be recommended for better specification of breast lesions especially in women with dense breasts and for staging of breast cancer.

Keywords

• Artificial intelligence
• Contrast-enhanced mammography
• Early detection of cancer
• Breast cancer
• Mammography
• Breast cancer screening