Thyroid Ultrasound Image Segmentation Based On Phase Geometric Active Contour Model

According to World Health Organization statistics, thyroid disease has become one of theworld’s top five causes, expected will jump to the second largest cause in2020. Ultrasonicdiagnosis becomes the preferred method of thyroid disease inspection, due to its low-cost,non-destructive, repeatability, high sensitivity advantages. Due to the impact of medicalworkers’ diagnosis level and the visual fatigue, the diagnostic outcome is subjective, and thediagnostic process is laborious and time consuming. Therefore, to achieve thyroid ultrasoundcomputer-aided diagnosis is necessary. It can provide the necessary disease conditioninformation for medical workers to improve diagnostic accuracy, reduce the workload ofmedical workers. Thyroid ultrasound image segmentation can segment the lesion area of thethyroid, is to achieve an important part of computer-aided diagnosis.Based on speckle noise, low contrast, inhomogeneity of the ultrasound image, geodesicactive contour models (GAC), one of geometric active contour models, has been improved.The geometric active contour model based on the phase information (PCGAC) has beenproposed. The model can segment the boundary contour of thyroid nodules. The main work inthis paper is as follows:This GAC model based on gradient boundary is sensitive to noise and does not have theability to detect weak edges. The Phase Congruency edge detection algorithm has bettercapability to detect noise images and weak edge. The paper constructs the stop function basedon the phase information, instead of the GAC model’s edge stop function. The geodesic activecontour segmentation model based on the Phase Congruency (PCGAC) has been received.The segmentation results explains the PCGAC model based on Phase Congruency is betterthan traditional GAC model, accuracy has been greatly enhanced.