Clinical–Alimentary TractArtificial Neural Network as a Predictive Instrument in Patients With Acute Nonvariceal Upper Gastrointestinal Hemorrhage
Section snippets
Setting and Study Design
The study was divided into two parts. The first part consisted of the development, training, and subsequent validation of an ANN for prediction of outcome in patients with acute nonvariceal UGIH at the primary study institution, University Hospitals Case Medical Center (UHCMC), Cleveland, Ohio. We also directly compared ANN with the Rockall scoring system in this internal cohort. The second part of the study was for external validation of the ANN in a different patient population and for direct
Results
During the period from January 1998 to June 1999, a total of 387 patients (195 men [50.4%], mean age [SD] 66.4 [± 17.2] years) were admitted to the primary study institution (UHCMC) with acute nonvariceal UGIH. All data were prospectively collected; a random sample of 10% of all patient records was reevaluated by another investigator (FF) to check accuracy of data extraction. The etiologies of nonvariceal UGIH were as follows: duodenal ulcer in 129 (33.3%), gastric ulcer in 99 (25.6%), erosive
Discussion
In acute lower GI hemorrhage, we have shown that ANN can accurately predict patient outcome and is superior to conventional multiple logistic regression when tested in an independent, external patient population.18 In our current study in acute nonvariceal UGIH, we have shown that ANN (a nonendoscopic method) performed at least as well as the complete Rockall score (an endoscopic method) in predicting major SRH and need for endoscopic therapy, even in an independent external patient population.
References (37)
- et al.
Acute upper GI bleeding: did anything change?Time trend analysis of incidence and outcome of acute upper GI bleeding between 1993/1994 and 2000
Am J Gastroenterol
(2003) - et al.
Acute gastrointestinal bleeding
Med Clin North Am
(2000) - et al.
Nonvariceal upper gastrointestinal bleeding: epidemiology and diagnosis
Gastroenterol Clin North Am
(2005) - et al.
Endoscopy-based triage significantly reduces hospitalization rates and costs of treating upper GI bleeding: a randomized controlled trial
Gastrointest Endosc
(1999) - et al.
Outpatient management for low-risk nonvariceal upper gastrointestinal bleeding: a randomized controlled trial
Gastrointest Endosc
(2002) - et al.
Selection of patients for early discharge or outpatient care after acute upper gastrointestinal haemorrhage
Lancet
(1996) - et al.
Incremental value of upper endoscopy triage of patients with acute non-variceal upper gastrointestinal hemorrhage
Gastrointest Endosc
(2004) - et al.
Prediction of outcome of acute GI hemorrhage: a review of risk scores and predictive models
Gastrointest Endosc
(2004) - et al.
Introduction to neural networks
Lancet
(1995) Application of artificial neural networks to clinical medicine
Lancet
(1995)
Analysis of clinical variables driving decision in an artificial neural network trained to identify the presence of myocardial infarction
Ann Emerg Med
Prediction of outcome in acute lower gastrointestinal haemorrhage based on an artificial neural network: internal and external validation of a predictive model
Lancet
Variation in outcome after acute upper gastrointestinal haemorrhage
Lancet
Prospective validation of the Baylor bleeding score for predicting the likelihood of rebleeding after endoscopic hemostasis of peptic ulcers
Gastrointest Endosc
A prospective outcome study of patients with clot in an ulcer and the effect of irrigation
Gastrointest Endosc
A risk score to predict need for treatment for upper-gastrointestinal haemorrhage
Lancet
Trends in management and outcomes of acute nonvariceal upper gastrointestinal bleeding: 1993–2003
Clin Gastroenterol Hepatol
Epidemiology of hospitalization for acute upper gastrointestinal hemorrhage: a population-based study
Am J Gastroenterol
Cited by (50)
Application of Artificial Intelligence to Gastroenterology and Hepatology
2020, GastroenterologyCitation Excerpt :ML models have therefore been developed to identify patients at risk for recurrent bleeding and those most likely to require treatment, and to estimate mortality. These models use clinical and/or biological data and identify these patients with approximately 90% accuracy.113–117 An ML model, developed in a retrospective analysis of 22,854 patients with peptic ulcer and validated in 1265 patients, was able to identify patients with recurrent ulcer bleeding based on their age, level of hemoglobin, gastric ulcer, gastrointestinal diseases, malignancies, and infections.
Previous Use of Antithrombotic Agents Reduces Mortality and Length of Hospital Stay in Patients With High-risk Upper Gastrointestinal Bleeding
2019, Clinical Gastroenterology and HepatologyDisease-free survival assessment by artificial neural networks for hepatocellular carcinoma patients after radiofrequency ablation
2017, Journal of the Formosan Medical AssociationModeling of glucose release from native and modified wheat starch gels during in vitro gastrointestinal digestion using artificial intelligence methods
2017, International Journal of Biological MacromoleculesCitation Excerpt :Based on the literature review, the ANNs had been a precise and appropriate instrument for modeling of many biological/physiological systems (in vitro and in vivo). For instance, ANNs were applied for predicting stigmata of recent hemorrhage [18], glucose monitoring in the blood and cerebrospinal fluid [49] and prediction of outcome in acute lower-gastrointestinal hemorrhage [50]. Hybrid GMDH-type neural network was developed for estimation of the AGR during the simulated gastrointestinal digestion condition.
Artificial neural network predicts the need for therapeutic ERCP in patients with suspected choledocholithiasis
2014, Gastrointestinal Endoscopy
The authors thank Douglas B. Haghighi, DMD, MD, and Judith A. Gonet, RN, for their assistance in data collection and entry.
- 1
Dr Wong’s research was supported by a 2003 ASGE, Wilson-Cook Endoscopic Research Career Development Award.