| Objective:Part ?: The reflux symptom index scales of the patients who were treated in the Otolaryngology Department of the Second Hospital of Jilin University for throat discomfort are evaluated in this study,so as to analyze whether the patients have laryngopharyngeal ref lux related symptoms and disease characteristics.Part ?: Artificial intelligence is used to extract and classify th e characteristics of electronic laryngoscope images between the experimental group and the control group and recognize abnormal images.It provides object evidence for the early screening of laryngopharyngeal reflux related symptoms,and lays the foundation for diagnosing laryngopharyngeal reflux disease accurately and objectively.Methods:Part ?:The patients received in the Otolaryngology Department of the Second Hospital of Jilin University from October 2018 to December 2019 for throat discomfort were selected and instructed to fill in the simplified Chinese version of RSI scales.The group with RSI>13 was classified to RSI positive group,while the group with RSI ? 13 was classified to RSI negative group.A total of 936 valid scales were collected.Based on the statistical analysis of various data of clinical characteristics in the scales,the occurrence rate and clinical characteristics of laryngopharyngeal reflux related symptoms of the patients received in the Otolaryngology Department of the Second Hospital of Jilin University for throat discomfort were summarized.Part ?: 939 electronic laryngoscope images of patients who received regul ar pre-operation electronic laryngoscope exams in the Thyroid Surgery Department and Otolaryngology Department with RSI ? 13 from October 2017 to December 2019 were collected as the control group.2,489 electronic laryngoscope images of patients who were tre ated in the Otolaryngology Department of the Second Hospital of Jilin University with RSI>13 were collected as the experimental group.The characteristics of laryngoscope images between the two groups were extracted and classified with the computer algorith m of artificial intelligence.On the basis of images grayscale and textual features,the characteristics of electronic laryngoscope images were distinguished,and the laryngeal injury caused by reflux early was accurately identified,which provided patholo gical warning for clinicians.Results:Part ?:1.A total of 936 valid RSI scales were collected,including 562 scales for RSI>13 and 374 scales for RSI ? 13.With RSI>13 as the standard,the RSI positive rate was 60.04% among all the patients in Otolaryngology Department of the Second Hospital of Jilin University for throat discomfort.2.Among all patients in three age groups(? 44 age,45~ 59 age,? 60 age),the RSI positive rate was 63.55%,62.90% and 57.56%.The comparative differences among the g roups had no statistical meaning(?2=3.305,P=0.192).The male patients' RSI positive rate was 57.57%,and the female patients' RSI positive rate was 52.00%.The comparative differences among the groups had statistical meanings(?2=4.376,P=0.0364).3.In the 9 symptoms included in RSI scales,all the frequency of occurrence of each symptom in RSI positive group was obviously higher than in RSI negative group,with statistic difference(P<0.001).The highest frequency of occurrence of symptom in RSI positive group was continuous clean throat(92.53%),followed by pharyngeal foreign body sensation(88.79%)and hoarseness or dysphonia(74.91%).The lowest frequency of occurrence of symptom was dysphagia,only 31.32%.The rank of typical symptom of gastroesophageal reflux of heartburn,chest pain and stomachache was only 5 and the frequency of occurrence was 65.48%.4.In the positive group of RIS,the highest score in all symptoms,namely the first three of the severity degree in turn was pharyngeal foreign body sensation(3.33±2.01),continuous clean throat(3.04 ±1.96),and hoarseness or parathria(2.34 ±2.05).Among them,the slightest symptom was dysphagia(0.58±1.26).5.The total score of RSI was related to all symptoms.The continuous clean throat was of strong correlation(r=0.513).The two symptoms(r=0.503)between the cough after eating or lying down and the annoying cough were of strong correlation(r>0.5).Part ?:1.This research extracted the characteristics of electron ic laryngoscope images between experimental group and control group with artificial intelligence.The result showed that there were differences in grayscale distribution and textual features on the whole between the electron ic laryngoscope images of two groups.The comparison between eigenvalue groups had statistical meanings(P<0.001).2.Four sorting algorithms of linear discriminant analysis,K-nearest,decision-making tree and na?ve bayes were used on the basis of grayscale characteristics,textural features and grayscale characteristics+textural features,to classify the electron ic laryngoscope images between the experimental group and the control group.The KNN discriminant algorithm had high accuracy in the classification o f grayscale characteristics and reached 85.33%;liner sorting algorithm had high accuracy in the classification of textural features and reached 88.63%;decision-making tree sorting algorithm had high accuracy in the classification of grayscale and textura l comprehensive features and reached 98.01%.3.There were statistical differences in arytenoid mucosa red index of electron laryngoscope images between the two groups(P=0.024).There were obvious statistical differences in postcommissure mucosa black and wh ite pixel value(P=0.022).Conclusion:1.The positive rate was 60.04% of patients received in the Otolaryngology Department of the Second Hospital of Jilin University for throat discomfort.Three most common symptoms were continuous clean throat,pharyngeal foreign body sensation and hoarseness o r dysphonia.The RSI score had strong correlation with continuous clean throat.It can be seen that the occurrence of laryngopharyngeal reflux is related to gender and has no connection with age.2.There were obvious difference of electron ic laryngoscope ima ges in gray distribution and texture features between the experimental group and the control group.The classification accuracy of the electron ic laryngoscope images by sorting algorithm is up to 98.01%.On this basis,abnormal images can be identified earl y to provide objective reliable evidence for LPRD early screening and lay foundation for later research. |
PDF Full Text Request