The Application Value Of Convotional Neural Network For Accurate Diagnosis Of Skull Base Fracture

Research background and purposeWith the development of society,the incidence of craniocerebral injury,especially heavy craniocerebral injury,is increasing.Due to the complex anatomical structure,bone curvature and scanning level differences of skull base bone,low-seniority radiologists are easy to cause missed diagnosis and misdiagnosis,resulting in clinical diagnosis delay and medical disputes when working independently.In recent years,numerous applied studies of deep learning in medical imaging show obvious advantages in image segmentation,classification and feature recognition,playing an important role in medical-assisted diagnosis.The Convolutional Neural Network has excellent robustness and high accuracy compared to other AI algorithms in the image processing work.This paper discusses the application value of convolutional neural network in the diagnosis of CT fracture by studying Tiny Net-CNN model on CT images of skull base fractures.Materials and MethodsFrom 2010 to 2019,skull CT images were collected from PACS system of many hospitals in Shenzhen and 2467 normal patients.After nano-rejection criteria and CNN model training,a total of 2488 skull base fractures and 1628 CT images of normal patients were included in the study.After manual fracture annotation of the CT images,the training and test sets were randomly assigned.Through CNN,skull area identification algorithm model and skull fracture detection algorithm model were constructed.Subsequently,the model of skull base fracture area identification and skull fracture and skull base fracture in the test are accuracy,recall,F1 value,average diagnosis time consuming;and compare diagnostic efficiency with the test data of manual group(low seniority radiologist).ResultAfter the stable model obtained by CNN calculation,the results showed that the fracture,anterior,middle and posterior skull base accuracy were <0.5,lower than the manual group(both> 0.63);the recall> 0.89,better than the manual group(all <0.8);the average diagnosis time was(3.12±2.67)s,less than the manual group.In the skull base fracture area test,accuracy: anterior skull base> middle skull base> posterior skull base,recall: middle skull base> posterior skull base> anterior skull base.ConclusionThe algorithm model of skull base fractures based on CNN is superior to the artificial test results in recall rate and diagnosis time consumption for CT diagnosis of skull base fractures in patients with craniocerebral trauma,which has certain value in assisting clinical diagnosis,reducing missed diagnosis and diagnosis time consumption.