Research On Artifact Suppression Method Of Optical Coherence Tomography

Optical coherence tomography(OCT)is a new type of medical imaging technology,and its endoscopic application is gradually developing in recent years.Due to factors such as the complexity of the scanning part and the friction of the driving motor,the speed of the endoscope is uneven during the imaging process,and motion artifacts are generated in the final imaging.To suppress the artifact problem in OCT endoscopic imaging and improve the imaging quality,two algorithms for suppressing motion artifacts are proposed in this thesis.First,a feature extraction-based image registration algorithm is proposed to suppress the motion artifact problem.Artifact suppression is achieved by aligning the scanned frames with image registration.This method mainly improves the feature extraction strategy of the FAST algorithm according to the characteristics of the OCT frame data in the endoscope,which improves the quality of the extracted feature points and ensures the extraction efficiency of the original algorithm.In the evaluation of artifact results,using MSE evaluation,the result is improved by up to 10.1%,and using SSIM evaluation,the result was increased by up to 7.2%.To suppress motion artifacts further in endoscopic OCT imaging and meet the realtime requirements,this thesis proposes a new unsupervised image registration model based on the U-Net architecture and using large convolution kernels to improve the feature extraction and restoration capabilities of the encoder and decoder.Through experimental analysis,the model is able to achieve a processing time of around 100 ms.In addition,the proposed method achieved a maximum improvement of 25.1% in PSNR and 23.7% in SSIM for artifact suppression.This thesis explores methods to suppress motion artifacts in OCT imaging under endoscopy,and the proposed method not only achieves excellent artifact suppression results,but also makes significant progress in algorithm processing speed.