Efficient Classification Of Benign And Malignant Thyroid Tumors Based On Characteristics Of Medical Ultrasonic Images

As the health issue is being concerned by more and more people, the workload of a clinical doctor becomes larger. As there are many images to diagnose every day for a medical ultrasonic doctor, image pattern recognition and classification technologies for medical ultrasonic images are necessary to reduce the workload of the clinical doctor. The major image pattern recognition methods include Bayesian pattern classifier, support vector machine method, and neural network model. These image pattern classification methods present good image classification performance but require large training dataset and long training time. As such, efficient characteristics-based image pattern classification methods were discussed in this paper, standard deviation classification method and aspect ratio classification method. They were applied to the recognition and classification of benign and malignant thyroid tumors on medical ultrasonic images. These image pattern recognition methods were built upon the inherent characteristics of the benign and malignant thyroid tumors which were presented on the ultrasonic images. The efficient classification methods interpreted in the paper demonstrated good classification performance, which verified that the characteristics-based image pattern classification methods can be utilized in effective image classifier construction.