AI-assisted diffuse correlation tomography for identifying breast cancer

• Published in: Journal of biomedical optics Vol. 30; no. 5; p. 055001
• Main Authors: Zhang, Ruizhi, Lu, Jianju, Di, Wenqi, Gui, Zhiguo, Wan Chan, Shun, Yang, Fengbao, Shang, Yu
• Format: Journal Article
• Language: English
• Published: United States Society of Photo-Optical Instrumentation Engineers 01.05.2025; SPIE
• Subjects: Adult; Aged; Algorithms; Artificial Intelligence; Breast - blood supply; Breast - diagnostic imaging; Breast cancer; Breast Neoplasms - blood supply; Breast Neoplasms - diagnostic imaging; Developing countries; Female; Health aspects; Health care industry; Humans; Image Interpretation, Computer-Assisted - methods; Image Processing, Computer-Assisted - methods; International economic relations; Medical imaging equipment; Medical screening; Middle Aged; Mortality; Rankings; Taiwan; Tomography - methods; Taiwan; clinical imaging; breast cancer; diagnosis; artificial intelligence; microvascular blood flow
• ISSN: 1083-3668; 1560-2281; 1560-2281
• DOI: 10.1117/1.JBO.30.5.055001

Diffuse correlation tomography (DCT) is an emerging technique for the noninvasive measurement of breast microvascular blood flow, whereas its capability to categorize benign and malignant breast lesions has not been extensively validated thus far, due to the difficulties in instrumentation, image reconstruction algorithms, and appropriate approaches for imaging analyses. This artificial intelligence (AI)-assisted DCT instrumentation was constructed based on a unique source-detector array and image reconstruction algorithm. The DCT images of breasts were obtained from 61 females, and AI models were utilized to classify breast lesions. During this process, the blood flow images were either extracted as feature parameters or as global inputs to the AI models. As the validations of DCT instrumentation, the blood flow images obtained from longitudinal monitoring of healthy subjects demonstrated the stability of DCT measurements. For patients with breast diseases, comprehensive analyses yield an AI-assisted classification with excellent performance for distinguishing between benign and malignant breast lesions, at an accuracy of 97%. The AI-assisted DCT reflects functional abnormalities that are associated with cancellous-induced high metabolic demands, thus demonstrating the great potential for early diagnosis and timely therapeutic assessment of breast cancer, e.g., prior to the tumor formation or proliferation of microvascular networks.