Design And Implementation Of Cardiac Coronary Artery Calcification Detection System Based On Deep Neural Network

Coronary artery calcification is a common arterial disease and is the main cause of coronary heart disease.In order to avoid the occurrence of medical accidents,the use of scientific and technological means for auxiliary diagnosis and treatment has been widely used in various medical fields.At present,most of the studies on adjuvant treatment of calcification are based on ordinary computer tomography images,and there are very few studies based on coronary computed tomographic angiography.However,angiography has become the most extensive clinical diagnosis basis due to its ease of shooting and less harm to patients.Compared with CT images,calcification has lower contrast and similar morphology to the background environment in angiography,which has higher identification difficulty.Therefore,the study of calcification lesions based on angiography is both challenging and significant.This thesis is devoted to researching and designing suitable detection algorithms,and builds a set of calcification detection auxiliary diagnosis system with practical application value based on coronary angiography.First,this thesis establishes the coronary angiographic-based calcification lesion dataset,which is the largest dataset known to this thesis.The dataset has good universality and provides consistent usage with public dataset.Secondly,this thesis demonstrates the feasibility and rationality of using artificial intelligence technology to detect calcification lesions on coronary angiography.A series of verification experiments are carried out based on the current cutting-edge algorithms in the detection field.The experimental results show that the use of convolutional neural network can overcome the difficulties of lack of color in angiography images and unclear visual features of the target area,effectively distinguish the calcification lesion area from the similar background area,and locate the lesion area more accurately.Finally,based on the observation of the developing state of calcification lesions on angiography images,this thesis conducts targeted optimization design and processing of the model structure and original dataset,proposes a detection algorithm suitable for this research scene.The optimized model proposed in this thesis can detect calcification lesions with an accuracy of 83.82%,a recall rate of 83.58%,and a comprehensive index F1 Score of 83.70%.Compared with other mainstream detection models,all evaluation indicators of the model proposed in this thesis have been greatly improved.At the same time,there is consistency between the lesion detection results of the optimized model and the clinical diagnosis results.