Abstract P6-01-02: Machine learning-based structural analysis and oxygen saturation measurement of tumor-associated vessels in breast cancer using a photoacoustic tomography system

Abstract

Introduction Breast cancer induces angiogenesis, one of the primary factors responsible for tumor progression. Therefore, the ability to visualize angiogenesis at a higher resolution is crucial. Photoacoustic tomography is a noninvasive method of visualizing angiogenesis involving light absorption and ultrasonic wave emission. If the irradiation light wavelength is adjusted for hemoglobin, vascular imaging is possible. Furthermore, using two wavelengths for oxidized and reduced hemoglobin, “S-factor,” can be calculated, which nearly corresponds to oxygen saturation. Therefore, photoacoustic imaging allows the assessment of breast lesions from vascular structural and functional viewpoints. Objectives This study aimed to demonstrate the possible utility of photoacoustic tomography for clinical application focusing on the morphologic features and oxygen saturation status of breast tumor-related vessels. Methods For the morphological analysis, we applied a machine learning-based method for automatic vessel extraction, and for the functional analysis we evaluated hemoglobin oxygen saturation calculating signals obtained at two wavelengths. In our system, a 3D ultrasound image was simultaneously acquired as a volume image of a tumor, which helped analyze the positional relationship between the vessels and the tumor. Results On morphological analysis, the fine structure of tumor-related vessels was rendered in high resolution. In our system, the blood vessels branched toward the tumor 2-3 more times more frequently than observed on contrast-enhanced MRI, illustrating a finer level of blood vessels near the tumor on our system than on MRI. Next, we analyzed the six morphologic features of vessels (radius, volume, curvature, contraction, maximum angle and vessel branch number) that are associated with the pathologic condition in neuroscience. We determined that the feature distribution of vessels located close to the tumor differed from that located away from the tumor. For example, vessels near the tumor had higher curvature, which means they are more tortuous than healthy vessels. The difference in the distribution of all six features was statistically significant on the Kolmogorov-Smirnov test. On functional analysis, S-factor measurement of the healthy human breast demonstrated clearly demarcated arteries and veins. The S-factor of any artery was nearly 100%, while that of the veins inside the breast cancer tended to be a little higher (approximately 5%) compared to that in the healthy part. This tendency of veins was not recognized in benign tumors. This could show arteriovenous shunt in cancer microenvironment. We found low saturation signals emerging in the tumor tissue following bevacizumab-containing chemotherapy, indicating the possibility that our system reveals microenvironment changes. Discussion If our system can identify the structure or oxygen saturation characteristics unique to tumor-associated vasculature, it could contribute to the improved accuracy of breast cancer diagnosis and allow the observation of tumor vessel normalization because of the drug treatment. An earlier grasp of the therapeutic effect could lead to the provision of individualized medicine. Citation Format: Matsumoto Y, Gu L, Bise R, Asao Y, Sekiguchi H, Yoshikawa A, Ishii T, Takada M, Kataoka M, Sakurai T, Yagi T, Sato I, Togashi K, Shiina T, Toi M. Machine learning-based structural analysis and oxygen saturation measurement of tumor-associated vessels in breast cancer using a photoacoustic tomography system [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P6-01-02.