Research And Development Of Pharyngeal Image-aided Diagnosis Based On Machine Learning And Its Acquisition Instrument

The pharynx is one of the important organs of the human body.When the virus enters the oral cavity,it can cause pharynx infection,and it can also cause trauma due to external factors.In severe cases,it can threaten the life of the patient.Therefore,medical examination of the pharynx is very important.Failure to cooperate with the examination caused by the patient's pharynx discomfort,or misjudgment by the medical staff due to fatigue,resulting in delay in treatment or misdiagnosis.In order to effectively solve this problem,this article develops a pharynx image acquisition instrument and its image-assisted diagnosis system.This device takes pictures or photographs of the patient's pharynx through handheld operation.The data is transmitted via wireless network and can be immediately consulted by professional dentists.,It can also transmit data to the cloud server,and recognize the patient's pharynx images through machine learning,assisting medical staff in judging the condition more quickly and accurately.Based on the above ideas,the research content of this paper is as follows:(1)Development of the pharyngeal image acquisition instrument.According to the requirements of pharyngeal image acquisition,combined with the analysis of product audiences,the overall design idea of the pharyngeal image acquisition instrument is determined to be compact and easy to carry.The image sensor,lighting source and main control chip are analyzed and selected,and the data transmission is determined.mode.(2)Aiming at the detection of the overexposed area in the pharyngeal image,an adaptive segmentation algorithm based on the HSI color space is used to make the overexposed area fully segmented.In the HSI color space,the use of adaptive segmentation for the components S and I can detect the overexposed area,and the corrected image can be obtained by filling and repairing.Finally,the quality of the image is evaluated,and the results show that the algorithm is reliable and effective in restoring the texture,color and other information of the overexposed area.(3)Aiming at the detection of the underexposed area of the pharynx image,a multi-scale Retinex improved algorithm based on HSV color space is proposed to repair the texture and color information of the front exposure area of the image.The image quality evaluation experiment on underexposed images shows that the algorithm can improve the brightness and saturation of the image under the condition of effectively retaining the same hue,which proves the effectiveness of the algorithm in this paper.(4)An image-aided diagnosis model of the pharynx acquisition instrument based on deep learning,which consists of three identical neural network sub-models and a Bagging fusion model.It includes 5 layers of convolutional layer and pooling layer.After each convolutional layer is normalized,after the final pooling layer,a one-dimensional vector is output.Use Bagging to feature fusion of the three model outputs.Finally,through experimental verification,the recognition accuracy of pharynx lesions is 93.25%,which proves the effectiveness of this model.