Assessment of cancer therapy effects using texton-based characterization of quantitative ultrasound parametric images

This paper proposes the application of texton-based approach for textural characterization of quantitative ultrasound parametric maps, in order to assess noninvasively the progressive effects of cancer treatment in preclinical animal models. Xenograft tumour-bearing animals were treated with chemotherapy. Ultrasound data were acquired from tumours prior to, and at different times after exposure, and quantitative ultrasound spectral parametric maps were generated. Texton-based features were extracted from 0-MHz Intercept parametric maps and applied to differentiate between preand posttreatment states. The classification error was then translated into a quantitative measure of the treatment effects. Obtained results demonstrated a very good agreement with histological observations, and suggested that the proposed approach can be used noninvasively to evaluate the progressive effects of cancer treatment.