Research On Three-dimensional Reconstruction And Visualization Of Medical Ultrasonic Images

The main research content of this paper is the three-dimensional reconstruction and visualization methods of medical ultrasound images.Ultrasound imaging plays an indispensable role in clinical diagnosis due to its repeatability,convenience,nonradiation,and low price.Especially for pregnant women and babies,ultrasound imaging is the preferred medical imaging check-up method.Therefore,ultrasound imaging is extremely important for the early prevention and diagnosis of diseases.Doctors can directly obtain the cut surface images by the traditional ultrasound imaging technology.Doctors need to infer the three-dimensional overall information of the tissue,organ or the diseased area based on the two-dimensional local information contained in the cut surface,and then give the disease assessment results and treatment plan accordingly,which often relies on doctors’ clinical experience.It is difficult to accurately determine the location,size,geometric shape and spatial relationship between the lesion and surrounding tissues and structures based on the subjective analysis of doctors alone.Therefore,it is meaningful to perform three-dimensional reconstruction on two-dimensional ultrasound images so that doctors can see the whole picture of tissues and organs intuitively.Usually there is no corresponding 3D model for ultrasound image sequences.Therefore,the previous 3D reconstruction work of ultrasound images only ends with the reconstructed 3D model,and the evaluation of the 3D reconstruction effect is lacking.In response to this problem,this paper proposes a U-shaped 3D reconstruction and visualization method simulation and performance evaluation framework,which can be used to compare the performance of 3D reconstruction methods.The threshold value of the classical isosurface extraction surface rendering algorithm is often difficult to determine.This paper proposes a modified Marching Cubes algorithm based on foreground extraction.This algorithm allows the threshold value to be arbitrarily selected in a large range through image processing,which improves the flexibility of the progressive cube algorithm.The classic Parallel Contouring algorithm often has branching problems.This paper proposes an improved Parallel Contouring algorithm for the key points of the included angle,which can effectively solve the branching problem.Aiming at the problem that there are few similarity evaluation methods for three-dimensional models that meet the requirements of rotation invariance and scale invariance,this paper proposes a similarity evaluation algorithm based on singular value decomposition and geometric feature inverse sampling.This paper uses the above-mentioned three-dimensional model similarity evaluation method to compare the performance of the above two improved three-dimensional reconstruction methods on the above U-shaped frame,and finally uses the better performance method to experiment on the real ultrasound breast image sequence.