Automatic Detection And Image Enhancement Of Hyperparathyroidism

The parathyroid gland is endocrine gland.Its major function is to regulate blood calcium content and metabolize calcium and phosphorus in the body.When the parathyroid gland is abnormal,it has a great impact on the health of the body,which may lead to hypocalcemia convulsions or cardiomyopathy,high blood pressure and other serious diseases,and even threaten the life of the patient.At present,its most common medical diagnostic method is B-mode ultrasound.Parathyroid disease is not a rare disease,but there is a large gap between urban and rural medical conditions in China.Therefore,some first-line clinicians in primary hospitals generally have insufficient knowledge of such diseases,resulting in patients missing the best treatment opportunity due to misdiagnosis.There are two main problems in the clinical diagnosis process.One is that the position of the parathyroid gland cannot be accurately determined,and the other is that the definition of the corresponding pathological site is relatively low,which is not easy to distinguish.Therefore,this paper studies the computer aided diagnosis method for hyperparathyroidism,including detection part and image enhancement part,to solve two problems respectively.Determining the position of parathyroid glands in ultrasound images uses a object detection algorithm based on Faster R-CNN.This method is based on a large number of labeled data set,but there is currently no public source on the Internet.To solve this problem,this subject cooperates with the China-Japan Friendship Hospital.The data set is collected by professional doctors.The author produces a special data set format for the object detection algorithm,and establishs a standard data set for hyperparathyroidism detection This paper improves the Faster R-CNN algorithm from the perspective of network model structure and parameter optimization.Aiming at the variable shape of hyperparathyroidism and other tissue and organ noise problems,this paper proposes a object detection method suitable for parathyroid ultrasound images,which uses deformable convolution to improve the traditional convolution method,and introduces difficult sample mining technology.Verification experiment results show that the proposed algorithm has higher detection accuracy.Improving the clarity of the corresponding pathological parts using the image enhancement algorithm based on CLAHE.In view of the information loss caused by the CLAHE algorithm,this paper proposes a hybrid y-CLAHE algorithm,which combines the Gamma transform and the high pass detail enhancement algorithm,introduces human-computer interaction,and specifies region of interest for enhancement.The verification experiment results show that the enhanced image clarity and visual quality are significantly improved,especially the required local lesion area,namely the parathyroid gland.This method can assist the clinical diagnosis of parathyroid glands and improve the diagnosis rate.