Research And System Implementation Of Intelligent Diagnosis Algorithm For Chronic Gastritis Based On Electronic Nose Data

The incidence rate of chronic gastritis is as high as 60%in the population.As the early symptoms of chronic gastritis are not obvious,many people don’t pay enough attention to this disease.Stomach cancer remains an important cause of cancer-related death worldwide.Chronic gastritis progresses to stomach cancer with the main stages being atrophy,intestinal metaplasia and dysplasia.Therefore,by screening for chronic gastritis so that patients can receive timely treatment,the risk of stomach cancer can be reduced.The main methods of screening for chronic gastritis include diagnosis endoscopic and pathological biopsy.Both of these screening methods are invasive and can cause further harm to the patient.In addition,the test results need to be judged by professional doctors with his or her experience.However,the experience of doctors is not reliable.It can easily lead to missed diagnosis and misdiagnosis.Relevant studies have shown that chronic gastritis affects the human body odor.The odor can be collected by electronic nose devices.The electronic nose is now widely used in several fields and can provide a new perspective for intelligent diagnosis of chronic gastritis.Hence,this thesis uses an electronic nose device to collect data and combines it with artificial intelligence algorithms to achieve intelligent diagnosis of chronic gastritis.At present,there are few studies related to the collection of chronic gastritis data and the design of intelligent diagnostic algorithms using electronic nose devices at home and abroad.Meanwhile,the existing diagnostic algorithm models incorporate one-sided features and do not consider sample class-imbalance and overlap,making their predictive performance poor.In this thesis,we conduct research around electronic nose data as well as design and implement an intelligent diagnostic system for chronic gastritis.The main work and innovations are as follows:(1)Preprocessing and feature extraction of electronic nose data for chronic gastritis.As the chronic gastritis data collected by the electronic nose device has the problem of inconsistent acquisition length,it needs to be preprocessed.Furthermore,the original features,statistical features,frequency domain features and time domain features of the electronic nose data will be extracted and used for the training of artificial intelligence algorithms.(2)To address the problem of class-imbalance and overlap in the chronic gastritis data set collected by electronic nose,a diagnostic algorithm based on feature fusion and neighborhood negative sample elimination is proposed in this study.The experimental results on three pre-processed electronic nose datasets show that the proposed algorithm can effectively improve the diagnostic performance of chronic gastritis.(3)By processing the electronic nose data of chronic gastritis and training the model,we designed and implemented an intelligent diagnosis software system.The intelligent diagnosis software system can realize the effective management of electronic nose data.And It can predict and analyze the data of the electronic nose.The system can provide a new diagnostic assistance to doctors for chronic gastritis.