Identity Risk Factors Leading To False Negative Diagnosis Of The Capsule Endoscopy In Obscure Gastrointestinal Bleeding Patients And Evaluate A New Method For Assisting Capsule Endoscopy Diagnose

ObjectiveThe length of Small intestine is about5-7meter, it is a nomadic part of bowel which is packed by mesenterium and is away from both month and anus. Small intestine has its own feature such as its tortuous track and its length, all of which made it extremely difficult to identify small intestinal diseases. Thus, small intestinal diseases have always been a blind spot in the whole gut tract for diagnosis. Traditional exam methods including endoscopic and non-endoscopic methods.The former one including gastric endoscope, bowel endoscope, pushing endoscope and balloon endoscope, and non-endoscopic methods including ultrasound, radiological exam,CT,MRI and DSA. Though plenty of means to be chosen, the use of these methods are usually limited because of their own disadvantages and the special structure of small intestine. Thus, small intestine checking has always been a difficult problem for digestive doctors.The first capsule endoscopy (CE) was made by Given corporation in Israel at2001The coming out of CE has been view as milestone and open a new era in the history of Gastroenterology. CE used the movement of digestive tract itself as motivation, fulfilled a less suffering total small bowel exam. For the reason of its easy-operation, safety, CE has now become the most common use device for small bowel checking. From the day CE has coming out, investigators began their research on testing the safety and exam value of CE. For example, at2000,Applyyard and his team made an animal research, which showed that CE had a clear view of inner bowel ulcer, sub-mucosa tumor, parasite and foreign objects. Further investigation proved that comparing to pushing endoscope, CE had a similar positive rate, sensitivity and specificity, however, CE showed a high rate of total bowel exam rate. Thus, CE became wildly used in clinical field for its safety, sensitivity, specificity and well tolerance.Obscure gastroenterstinal bleeding(OGIB) was defined as ongoing or intermittent unclear gastrointestinal bleeding after both gastric endoscope and colon endoscope examination. Such kind of bleeding consist of5-10%of total digestive tract bleeding and was mainly caused by small intestinal disease. There were a lot of people did researches on CE regarding the diagnostic yield. It has been reported that the diagnostic yield of CE was between52%-88%, the positive diagnosis predictive value was greater than95%which is much higher than other traditional methods. Nevertheless, the negative result predictive value was80-86%which indicates a false negative result may be reported while operated CE. The use of CE indeed has the risk of raising false negative result which can be caused by plenty of reasons. For example, some special kind of disease may has an higher risk for generate false negative diagnosis and some inner features of CE itself made it incline to cause false negative diagnosis as well, even the tiredness of gastroenterologist while view CE pictures may lead to false negative diagnosis. However, there was few of study has been done concerning the false negative cases.In order to identify risk factors that may lead to false diagnosis, we design this study which retrospectively review patients with OGIB in Nanfang hospital, and analyzed susceptive risk factors that may lead to false negative diagnosis. Furthermore, on the purpose of understanding the risk of human-factor in leading false negative diagnosis, we collected pictures of false negative diagnosis cases and conducted retrospective picture-reading.By doing what mentioned about, we hope to provide some advices in clinical practice.Moreover, in order to simplify the reading process and reduce the rate of man-made false negative diagnosis. Our team developed a set of software which is to help screen invalid and repeat images. We use several ways to evaluate the efficiency and reliablitiy of this system, especially its application on assist diagnostic process. Firstly, we developed a score system which helps to build objective relation between quality of image and scores. Then, screening system was used and a respective image reading was proformed which gave each patient a score of their images after different screen conditions. At last, the changing of scores, image feartures, and ratio of disease remaining under different screen conditions were compared to evaluate the effect of the screening system.Methodology1. Data inclusion and exclusion criteriaTotal of616in-ward patients who did CE examination were collected in to our study as data pool,193of them were view as OGIB. Patient who meets the inclusion are put into positive diagnosis group and false negative diagnosis group respectively. Positive diagnostic group was defined as pathological lesions were found by CE which were comfirm by other examination method (e.g DSA,DBE or IOE) and/or more than2gastroenterologist. False negative group was defined as negative diagnosis of CE, however lesions were found and comfirmed by other methods. Patients who meet the following conditions were exclusion from the study1) nothing or susceptive lesion was found by CE, however no follow-up examination was conducted.2) insufficient data for analysis.2. Capsule endoscope exam and parameters settingAll patients used OMOM capsule endoscope and the results were analyzed by experienced doctors. Location of the lesions were based on the image features under CE:1)jejunum:Circle and gross folds with small interval, appears a flap-like overlap when constriction. Small blood vessels under mucosa.2) Circle but small folds, relative big blood vessels and folliculi lymphaticus present. Quality of the image is define as follows:1) good image:bowel preparation with clear sight of intestine cavity, moderate lightness or little intestinal contents but have no influence on diagnosis.2)bad image:poor sight for diagnosis because of improper brightness, too much intestine contents or think intestinal juice.3.Data collectionAccording to result of diagnosis, patients were divided into three groups, positive diagnostic group, false diagnostic group and undefined group respectively, and the constituent ratio was counted. The main aim of the study is to compare the differences between positive diagnostic group and false diagnostic group so as to identify risk factors which leads to false diagnosis. Several susceptive factors were considered and put into this study, such as demographic feature, disease features and CE features.11kinds of factors including age, sex, accompany diseases, capsule transmit time, type of purgative drug, history of disease, quality of CE picture,hemoglobin concentration, type of lesion and location of lesion. Among these factors, age hemoglobin concentration,transimit time of CE and history of disease were measurement date, rest of the factors were enumeration data.4. Retrospective image readingCollected all the images of the patients in the false negative diagnosis group, and picked up the same number of patients from the positive diagnosis group. All the image of the patients from these two groups were collected and analyzed by experienced doctors. The features of the images were compared between groups.5. Build and evaluate a novel image score system:Based on former work of our reseach group, the former image score system was advanced according to clinical practice experience and reports by other scholars. Then comparasion and evaluation of these two score system was operated.6. Build and evaluate a novel automatic image screen system: Cooperative work was down by our research team and department of medical engineering. An automatic image screen system was developed and then we chose25cases from database of capsule endoscopy and different screen conditions were used to screen their images, after the screening, several aspects including total score, picture features, processing time and rate of disease remaining were compared. By doing thing, we hope to understand the application value of this screen system and give more information for advance this system.5. Statistic methodAll the measurement data were analyzed through student’s t test, all the enumeration data were analyzed through chi square and partition chi-square test. Binary logistic regression was also used for analyze of the correlation between the factors and result of diagnosis. Repeat measurement data was analyzed by Repeated measures ANOVA.SPSS13.0was used for analyzing data.Resultl.The mean age of the positive diagnosed group was46.35±19.26years old, whereas in the false negative diagnosed group the mean age was37.89±19.06years old(p<0.05).Distribution of age has statistical difference between groups.2.Type and location of the disease, were found to have statistical differences between the groups.3.Image quality of the CE were statistically different between the two groups4.According to logistic regression, diverticulum disease indicate a higher risk of false4.negative diagnosis(OR=0.049, positive diagnosis group was set to be reference group), lower bowel diseases would have a much higher probability to be missed for diagnosis(OR=0.099when positive diagnose group was set to be reference).Poor quality of CE picture was another key factor that influenced the result of diagnosis, both chi-square test and logistic regression showed that poor quality of CE picture had negative influence on CE diagnosis(OR=0.162). Type of purgative agents also affect diagnosis, according to logistic regression, comparing to polyethylene glycol, sodium salt agent had a higher risk of leading to false negative diagnosis(OR=0.152)5.16%of the patients in the false negative group were cause by human-factor.6.The valid image ratio is much lower in the false negative group than the positive group(4.77%VS11.89%, t=5.993; p<0.001).Invalid image ratio is higher in the false negative group than the positive diagnosis group (43.57%VS16.61%,, t=6.675, P<0.001),repeat image ratio is lower in the false negative diagnosis group (51.65%VS71.49%,t=4.855, P<0.001).7, Score was related to the feature of the picture and can be used as a indicator of image quality:score made by system A was negatively related to valid image rate with r=0.576,95%CI (0.37,0.74), and positively related to invalid and repeat image rate with r=0.71395%CI (0.55,0.83) and r=0.62895%CI (0.43,0.78) respectively. While score system B was negatively correlated to valid image rate with r=0.576,95%CI(0.38,0.75), and positively related to invalid and repeat image rate with r=0.71495%CI (0.55,0.83) and r=0.62995%CI (0.45,0.79) respectively. The different between the two score system didn’t show statistic differences.8, Compare to score system A, system B was simpler and easier to operate. By using system B, we find that score between different diagnosis groups were statistic different, which is74.28±3.74VS68.91±7.11(T=3.169,p=0.003) repectively.9、By applying ROC curve, we identify score73.64is the most sensitive and special cut point for forecast and judge the reliability of a diagnosis.10、 Using different screen conditions of the screen system, we discovered that after screening process:total score (F=28.211,p<0.001) and valid image rate (F=74.22,p<0.001) were significantly increased;while total number of invalid images (F=57.52p<0.001),invalid image ratio(F=17.87,p<0.001)and repeat picture number (F=43.299,p<0.001)significantly reduce; however, ratio of repeat picture (F=1.154, p<0.330)、total available picture number (F=0.704,p=0.551) didn’t showed significantly change.11、Diagnose results may influence the effect of screening system. According to statistic analyze, different diagnostic groups showed different trend on valid ratio(F=3.86, p=0.01),invalid ratio(F=4.91, p=0.003) and repeat ratio change(F=4.62, p=0.004)12, By applying screen system, reading time can be significantly reduced (F=116.467, p<0.001).Though under the screen ratio of40%,20%and10%, the disease remain ratio was74.78%,62.13%,49.19%repectively and such change showed statistic significance(F=47.64, p<0.001).Disease discover rate remain100%.Conclusion1. Elder in age may help diagnosis diverticulum disease, lower small bowel disease,type of purgative agents and poor quality of CE pictures were risk factors that lead to false negative diagnosis.2. Man-made reason consist16%of the total false negative diagnostic rate.3. Comparing to diagnosed group, the false negative diagnosis group have a much lower rate of available pictures.(4.39%VS11.71%, p<0.001)4. Score made by the score system was closely related to the feature of pictures and can be used as a good indicator of the quality of the picture.It can also be applied to forecast and judge the reliability of a diagnosis.5. Image screen system can make a good clinical use. It can raise the total image score, moreover, the system reduce the total reading process time while keeping all the disease picture.