Early Evaluation Of Tumor Response To Chemotherapy Based On Ultrasound Spectrum Analysis From RF Signal

Cancer is one of three major diseases which seriously threatens human life in the world, and chemotherapy is the primary clinical treatment of malignant tumors currently. Early evaluation of tumor response to chemotherapy is of great significance for the treatment of tumor. The change of microstructures when and after tumor cell death has been an important target of early evaluation. Recent decades of research indicates that ultrasonic tissue characterization, achieving observing at sub-pixel level, is one of the most promising techniques to observe microstructures, which makes early evaluation of cancer chemotherapy efficacy possible. Aiming at the shortcomings of traditional cancer chemotherapy evaluation methods, this paper proposed to use ultrasonic backscatter signal RF spectrum for analysis, indirectly measuring the change of microstructures and evaluating chemotherapy efficacy at the early stage of treatment, which helps set out personalized cancer chemotherapy at early time. The paper mainly contains:1. This paper designs and develops a clinically application for early evaluation of the cancer chemotherapy efficacy and proposes a lossless, low-cost, noninvasive imaging method to achieve early efficacy evaluation.2. This paper adopts periodogram to estimate RF signal power spectrum. In order to reduce the impact of system parameters on result, the RF signal power spectrum from tumor tissue is normalize by the spectrum from the reflection plate, and then conduct linear fitting of the normalized spectrum to extract spectral feature closely related to the microstructures, including the slope, intercept, midband-fit, scatterer radius. Results show that the changing of tissue microstructures can be known via spectral characteristics.3. In order to further enhance the effectiveness of ultrasound back-scattered RF signals representing tissue characterization, a method based on AR model to estimate power spectrum and extract spectral feature is proposed. Meanwhile, a new characteristic values of mean scatter spacing estimated by AR cepstrum is increased. The results show that the spectral estimation and feature extraction based on AR model can reduce the variance of spectral features, which effectively reflects tissue microstructures.4. As for linear fitting normalized power spectrum, this paper adopts a least square method regression analysis and robust regression analysis. The results show the effectiveness of the two methods are basically the same according to analyzing the feature values of the fitting line.5. Based on the analysis above, this paper selects effective feature values to form a feature set, and using the standardized Euclidean distance to measure the similarity between the samples of the treatment group and the control group, and it shows that the intra-class similarity of samples from the two groups is low, while the inter-class similarity is high. The results show that the method of early evaluation of tumor response to chemotherapy based on ultrasound RF signal spectrum analysis is effective.The early response evaluation system we develop has been used to do some scientific research about evaluating tumor response to chemotherapy for two years by the tumor hospital of Sun Yat-Sen University.