Multi-variable quantitative ultrasound imaging for differential breast cancer diagnosis

Abstract

In this dissertation, a single-probe ultrasonic imaging system that captures multi-variable quantitative profiles is presented. As pathological changes cause biomechanical property variation, quantitative imaging has great potential for lesion characterization. The proposed system simultaneously extracts four clinically informative quantitative biomarkers, such as the speed of sound, attenuation coefficient, effective scatter density, and effective scatter radius, in real-time using a single scalable neural network. A b-mode contents-aware quantitative imaging scheme is proposed and enhances reconstruction accuracy by applying quantitative style information to structurally accurate b-mode images. In real-time medical imaging, spatio-temporal consistency is an important research challenge, as the temporal property variation hinders the accurate diagnosis of a lesion. Therefore, a residual quantitative imaging network is proposed and enhances spatio-temporal consistency of generated images by formulating a feed-back loop and inferencing residuals of adjacent frames. The performance of the proposed system was evaluated through numerical simulations and phantom, ex vivo and in-vivo measurements. In the phantom and ex vivo experiments, the SQI-Net demonstrated the classification of cyst, and benign- and malignant-like inclusions through a comprehensive analysis of four reconstructed images. In an in-vivo study, the proposed system demonstrates 0.907 area under curvature in the differential diagnosis of breast cancer