Research On Aortic Segmentation Algorithm Of CT Image Based On Adaptive Level Set Method

The aorta is the thickest arterial canal in the human body,which is responsible for pumping blood to various organs throughout the body.When the intima ruptures,the endometrium gradually separates from the endometrium under the impact of blood and forms a false cavity,this condition called aortic dissection.Aortic dissection is one of the more common cardiovascular diseases,and once it occurs,it is highly lethal if it is not treated urgently,which seriously endangers the life safety of patients.The main examination method for patients suspected of aortic dissection in hospitals is imaging examination,including computed tomography(CT),magnetic resonance imaging(MRI)and other techniques.The doctor then observes and analyzes the imaging data to identify the possible lesion area of the patient and make a diagnosis of the condition.Therefore,the segmentation of the aorta region based on CT images has a huge effect on the clinical diagnosis of doctors,but the examination results of a patient have hundreds of CT slides,which not only requires a lot of energy from experienced doctors to observe,but also affects the judgment of the condition due to factors such as doctors’ mistakes.In recent years,many models based on CT image aortic region segmentation have been proposed,which can provide advance indication of possible lesion CT slides,which has a great effect on doctors’ clinical diagnosis.Therefore,this paper proposes a CT image aortic segmentation algorithm based on level set method and deep learning,and the main work is(1)This paper analyzes the shortcomings of the traditional Distance Regularized Level Set Evolution(DRLSE)algorithm,and proposes an adaptive DRLSE algorithm based on image gray values,which improves the dependence of the traditional level set method on the parameter α,introduces the definition of internal and external detection domains,and designs an adaptive α calculation formula,and uses CT image grayscale information to guide the evolution of the level set function.Through the improvement of this algorithm,the DRLSE model can automatically determine the evolution direction and achieve stable evolution in the case of noise or multi-target.(2)Explore the reasons why the adaptive DRLSE algorithm is not suitable for CT data,process CT data by using morphological operations such as opening and closing operations,andintroduce scale factors,so that the model can change the evolution speed by itself through the image gray value during the evolution process,and an improved adaptive DRLSE algorithm is proposed,which enables the level set method to obtain more stable segmentation on CT data and reduce the occurrence of boundary leakage.(3)Based on the Unet network and the improved adaptive DRSLE algorithm,a complete aortic region segmentation process is designed.The Unet output result is transformed into the initial level set function,and the Unet rough segmentation result is optimized by using the level set method,and corresponding experiments are designed to verify the practicability and stability of the algorithm through the display and analysis of the algorithm process results.The experimental results show that after the optimization of the level set method,the segmentation effect is improved.