Differential Diagnosis Of Thyroid Nodules Based On Ultrasonic Radio-frequency Signals

The widespread use of ultrasonography(US)has contributed to the increased detection of thyroid nodules.Thyroid US depicts nodules in up to 70%of the population.However,less than 5%of these nodules are malignant.Malignant thyroid nodules have become one of the risks to human health in modem society.Medical ultrasound imaging has the advantages of real-time,nondestructive,cheap,good repeatability,high sensitivity,often used in the identification of benign and malignant thyroid nodules.At present,the gold standard for the identification of benign and malignant thyroid nodules is fine-needle aspiration biopsy(FNAB).The following method provides important information for clinical diagnosis to reduce the pain of patients and shorten the inspection period.A new method based on ultrasonic radio frequency signal was proposed to distinguish between benign and malignant thyroid nodules.The RF signal was analyzed in two domains(time domain and frequency domain),nonlinear analysis and back-propagation(BP)artificial neutral network.Firstly,select ultrasound radio frequency signals in the region of interest of thyroid nodules.In the time domain,the mathematical expectation(ME)of liver RF signal was extracted as the feature parameter.In the frequency domain,low-frequency wavelet coefficients mean(LWCM)and wavelet transform modulus maximum mean(WMMM)were extracted as the feature parameter.The largest Lyapunov exponent was as the feature parameter in nonlinear analysis.Finally,back-propagation(BP)artificial neutral network was employed to classify these RF signals.The accuracy rates with BP neural network are 95.1%in ME,92.7%in LWCM,97.6%in WMMM,and 82.9%in the largest Lyapunov exponent.The result showed that these feature parameters can successfully describe the features of thyroid nodule RF signal.This work provides a new idea for the diagnosis of thyroid nodules,and has important value in clinical application.