Assessment of Helicobacter pylori infection by deep learning based on endoscopic videos in real time.

BACKGROUND AND AIMS: Endoscopic assessment of Helicobacter pylori infection is a simple and effective method. Here, we aimed to develop a deep learning-based system named Intelligent Detection Endoscopic Assistant-Helicobacter pylori (IDEA-HP) to assess H. pylori infection by using endoscopic videos in real time. METHODS: Endoscopic data were retrospectively obtained from Zhejiang Cancer Hospital (ZJCH) for the development, validation, and testing of the system. Stored videos from ZJCH were used for assessing and comparing the performance of IDEA-HP with that of endoscopists. Prospective consecutive patients undergoing esophagogastroduodenoscopy were enrolled to assess the applicability of clinical practice. The urea breath test was used as the gold standard for diagnosing H. pylori infection. RESULTS: In 100 videos, IDEA-HP achieved a similar overall accuracy of assessing H. pylori infection to that of experts (84.0% vs. 83.6% [P = 0.729]). Nevertheless, the diagnostic accuracy (84.0% vs. 74.0% [P<0.001]) and sensitivity (82.0% vs. 67.2% [P<0.001]) of IDEA-HP were significantly higher than those of the beginners. In 191 prospective consecutive patients, IDEA-HP achieved accuracy, sensitivity, and specificity of 85.3% (95% CI: 79.0%-89.3%), 83.3% (95% CI: 72.8%-90.5%), and 85.8% (95% CI: 77.7%-91.4%), respectively. CONCLUSIONS: Our results show that IDEA-HP has great potential for assisting endoscopists in assessing H. pylori infection status during actual clinical work.

Factors associated with non-lifting of colorectal mucosal lesions.

BACKGROUND: Endoscopic submucosal dissection (ESD) is an effective treatment for colorectal tumors. However, lesions that cannot be lifted after submucosal injection are not indication for ESD. This is because the procedure is difficult, and the lesions are often considered as tumor invasion or submucosal fibrosis. The aims of this study are to evaluate the efficacy and safety of ESD for non-lifting lesions and to analyze the causes of non-lifting phenomenon. METHODS: This retrospective study included 29 patients with non-lifting colon lesions resected by ESD from February 2018 to September 2021. Cases were observed for demographics, endoscopic findings, treatment outcomes, adverse events and endoscopic follow-up. We studied the pathological features of lesions to explore the reasons for non-lifting. RESULTS: Among 29 cases of non-lifting lesions, 20 lesions (69.0%) were 30 mm in diameter or larger. Most of lesions (96.6%) were non-lifting in center, and only one lesions (3.4%) had non-lifting of one side. The en bloc and curative resection rates of ESD were 100 and 86.2%, respectively. There was one (3.4%) delayed bleeding, no perforations and other complications. No tumor recurrence occurred during the follow-up period. For pathological features, 16 (55.2%) non-lifting lesions had submucosal fibrosis and only 4 cases (13.8%) had deep submucosal invasion. There were 9 cases (31.0%) of non-lifting lesions due to musculo-fibrous of muscularis propria anomaly (MMPA). CONCLUSION: MMPA is another reason for non-lifting signs besides invasive carcinomas and submucosal fibrosis. ESD should be considered in patients with large non-lifting adenoma instead of surgery.

Correlation of the detection rate of upper GI cancer with artificial intelligence score: results from a multicenter trial (with video).

BACKGROUND AND AIMS: The quality of EGD is a prerequisite for a high detection rate of upper GI lesions, especially early gastric cancer. Our previous study showed that an artificial intelligence system, named intelligent detection endoscopic assistant (IDEA), could help to monitor blind spots and provide an operation score during EGD. Here, we verified the effectiveness of IDEA to help evaluate the quality of EGD in a large-scale multicenter trial. METHODS: Patients undergoing EGD in 12 hospitals were consecutively enrolled. All hospitals were equipped with IDEA developed using deep convolutional neural networks and long short-term memory. Patients were examined by EGD, and the results were recorded by IDEA. The primary outcome was the detection rate of upper GI cancer. Secondary outcomes were part scores, total scores, and endoscopic procedure time, which were analyzed by IDEA. RESULTS: A total of 17,787 patients were recruited. The total detection rate of cancer-positive cases was 1.50%, ranging from .60% to 3.94% in each hospital. The total detection rate of early cancer-positive cases was .36%, ranging from .00% to 1.58% in each hospital. The average total score analyzed by IDEA ranged from 64.87 ± 16.87 to 83.50 ± 9.57 in each hospital. The cancer detection rate in each hospital was positively correlated with total score (r = .775, P = .003). Similarly, the early cancer detection rate was positively correlated with total score (r = .756, P = .004). CONCLUSIONS: This multicenter trial confirmed that the quality of the EGD result is positively correlated with the detection rate of cancer, which can be monitored by IDEA. (Clinical trial registration number: ChiCTR2000029001.).

Factors associated with significant liver necroinflammation in chronic hepatitis B patients with cirrhosis.

We determined the association between various clinical parameters and significant liver necroinflammation in patients with chronic hepatitis B (CHB) related cirrhosis. Two hundred patients with CHB related cirrhosis were recruited in the final analysis. Clinical laboratory values and characteristics were obtained from the medical record. We performed analyses of the relationships between independent variables and significant liver necroinflammation by using binary logistic regression analysis and discriminant analysis. Significant liver necroinflammation (grade≥2) was found in 58.0% (80/138) of antiviral therapy patients and 48.4% (30/62) of non antiviral therapy patients respectively. Also, there were some significant differences in serum hepatitis B surface antigen (HBsAg), serum hepatitis B e antigen (HBeAg) and serum hepatitis B virus (HBV) DNA between antiviral therapy and non antiviral therapy patients. After that, aspartate aminotransferase (AST), total bilirubin (TBIL), total bile acid (TBA), prothrombin time (PT), aspartate aminotransferase to platelet ratio index (APRI) and serum HBV DNA were confirmed as independent predictors of significant liver necroinflammation in CHB patients with cirrhosis by univariate analysis and multivariate analysis (p = 0.002, 0.044, 0.001, 0.014, 0.01 and 0.02 respectively). Finally, receiver operating characteristic (ROC) curve analysis and discriminant analysis validated that these six variables together have strong predictive power to evaluate significant liver necroinflammation.

The characteristics and clinical outcome of drug-induced liver injury in a Chinese hospital: A retrospective cohort study.

The aim of this cohort study was to determine the characteristics and clinical outcome of 287 patients with drug-induced liver injury (DILI) in a Chinese hospital.Between January 2008 and January 2013, individuals who were diagnosed with DILI were selected. The complete medical records of each case were reviewed, and factors for the outcome of patients with DILI were extracted and analyzed using univariate and multivariate analysis.Two hundred eighty-seven cases identified as DILI were included in the study. A total of 105 different drugs were considered to be related to the hepatotoxicity. The main causative group of drugs was Chinese herb (n = 111). Liver failure developed in 9 (3.1%) patients, and 2 died (0.7%). Overall, complete recovery occurred in 92 (32.1%) patients. Univariate analysis and binary logistic regression analysis identified the digestive symptoms, jaundice, total bilirubin (TBIL), and direct bilirubin (DBIL) as independent factors for the non-recovery of DILI. Then the prediction model, including digestive symptoms, jaundice, TBIL, and DBIL, was built by using binary logistic regression analysis again. Receiver operating characteristic curve validated the strong power (area under the curve (AUC) = 0.907) of prediction model for predicting the DILI non-recovery.DILI is an important cause of liver test abnormalities, and Chinese herb represented the most common drug group. The factors such as digestive symptoms, jaundice, TBIL, and DBIL have effect on DILI outcomes. The prediction model, including digestive symptoms, jaundice, TBIL, and DBIL, established in this study is really an excellent predictive tool for non-recovery of DILI patients.

The combination of three molecular markers can be a valuable predictive tool for the prognosis of hepatocellular carcinoma patients.

Based on molecular profiling, several prognostic markers for HCC are also used in clinic, but only a few genes have been identified as useful. We collected 72 post-operative liver cancer tissue samples. Genes expression were tested by RT-PCR. Multilayer perceptron and discriminant analysis were built, and their ability to predict the prognosis of HCC patients were tested. Receiver operating characteristic (ROC) analysis was performed and multivariate analysis with Cox's Proportional Hazard Model was used for confirming the markers'predictive efficiency for HCC patients'survival. A simple risk scoring system devised for further predicting the prognosis of liver tumor patients. Multilayer perceptron and discriminant analysis showed a very strong predictive value in evaluating liver cancer patients'prognosis. Cox multivariate regression analysis demonstrated that DUOX1, GLS2, FBP1 and age were independent risk factors for the prognosis of HCC patients after surgery. Finally, the risk scoring system revealed that patients whose total score >1 and >3 are more likely to relapse and die than patients whose total score ≤1 and ≤3. The three genes model proposed proved to be highly predictive of the HCC patients' prognosis. Implementation of risk scoring system in clinical practice can help in evaluating survival of HCC patients after operation.

Fever of unknown origin: report of 107 cases in a university hospital.

Fever of unknown origin (FUO) is a challenging problem in clinical practice. Evaluation of patient's characteristics may illustrate the etiologies of FUO. In present study, 107 patients with FUO hospitalized in our inpatient department between 2010 and 2011 were investigated. The median age of the patients was 48 years (15-94). The median fever duration was 8.5 weeks (3-104). The median hospital stay was 8.5 days (1-51). Etiologies of FUO were identified as follows: infectious diseases 32 (29.9%), malignancies 19 (17.8%), inflammatory rheumatic diseases 18 (16.8%), and miscellaneous diseases 15 (14.0%). In 23 (21.5%) patients, the diagnosis remained unclear. Infection group had relative shorter average fever duration and hospital stay than other groups. Shortened mean fever duration was observed in geriatric age group. In conclusion, as the most common cause of FUO in the present study, infectious cases had relative shorter average fever duration and hospital stay, and geriatric patients had shortened average fever duration as well.