Automatic Alignment The Three Reference Standard Planes Of The Fetal Face From Three-Ultrasound Image System

Fetal facial ultrasound screening examinations is an important project during pregnant.Clinical commonly used three-dimensional ultrasonic surface pattern mainly provide fetal facial surface information, standard planes access still is the key aspect when using three-dimensional ultrasound scan fetal facial defects. But because of the uncertainty of the intrauterine fetal position, doctors need to be from the fetal facial three-dimensional ultrasound volume data manual checking fetal position, and then manually locate three orthogonal reference standard planes.This process depends on the doctor’s experience and time-consuming, thus reducing the efficiency of clinical diagnosis.Therefore, the paper brings forward a system for automatically aligning the relevance of the three reference scanning planes the fetal face in 3D ultrasound volumes, thus can reduce the workload of doctors,improve the standardization level of prenatal ultrasound.We take the facial three-dimensional ultrasound volume data as study object, to realize automatic aligning the relevance of the three reference scanning planes, the median sagittal plane, facial coronary plane and double eye axial plane.The paper puts forward the method the context constraint automatic detection of double eye, the first to automatic proofreading the median sagittal plane. First of all, this thesis defines the eye ball sphenoid bone model samples, the three dimensional space search series 9d(location, direction and size) parameters to the 4D(position and radius), greatly reduce the complexity and computation.Second, using 3D Haar-like to feature extraction, evaluation function feature selection and support vector machine(SVM) learning training to get the eye and sphenoid classifier. Finally, make full use of the sphenoid bone and eyeball space position relations invariance, eye and sphenoid exist in pairs, before and after feedback and multi-resolution analysis resolution strategy on-line detect the double eye, can effectively overcome the dark area influence on ultrasound images.Based on the median sagittal plane of automatic aligning, this paper propose through automatic detection face profile line, the linear equation for the face profile line, realize automatic aligning face coronal plane. Firstly, This thesis according to the location of the nasal bone determine the median sagittal plane of face profile area, based on the regionalsegmentation to get binary image, and getting hough straight line detection to get the face profile line about angle range, finally by the face profile template for the face profile line of the optimal fitting equation of the straight line. According to the nasal bone, double eye and face profile line to redefine the fetal facial three-dimensional space coordinate system to automatic aligning double eye axial plane.This study is based on VS2008, and by using VTK(Visualization Toolkit) achieved any three orthogonal views of 2D and 3D show, interested in interception of samples in the area.With testing 80 individual data(28 to 36 weeks), the experimental results show that our system with ultrasound doctors checked the median sagittal plane, coronal plane and horizontal transverse section three orthogonal reference standard aspect angle mean error are:3.836 ± 2.954(degree)、4.870 ±3.822(degree)、4.805 ± 4.005(degree),the distance mean error were 2.022±1.707(mm) 、 2.921 ±1.887(mm) 、 2.543 ±1.578(mm). Between users of the variability of further validation show that the automatic proofreading system can reduce the differences between users.Clinical workflow efficiency test, verify the automatic proofreading system can improve the efficiency of the doctor’s work. Therefore, this paper puts forward in facial proofreading system is a feasible and repeatable, and reliable method.The system constructed in this thesis is the first automatic alignment the three reference standard planes of the fetal face from three- ultrasound image system in related research area, it will fully computer technology combined with clinical application.