Can artificial neural networks predict which patients need a colonoscopy?
Gardiner, A; Maslekar, S; Duthie, Graeme
Introduction: Artificial neural networks (ANN) are computer programs used to identify complex relations within data sets undetectable with conventional linear statistical analysis. One such complex problem is the prediction of need for lower gastrointestinal endoscopy in individual patients consulting for gastrointestinal symptoms. Routine predictions have low accuracy and result in large numbers of normal colonscopies with obvious implications, both logistic and economic. We aimed to develop a neural network algorithm which can predict the need for lower gastrointestinal endoscopy in patients attending the routine outpatient clinics. MethodsProspective clinical data of 200 patients undergoing elective colonoscopy were collected. The specifically developed questionnaire included 40 variables based on clinical features. Complete data sets of 50% of the series were used to train the ANN: remaining 50% used for internal validation. The primary output was a positive finding on the colonoscopy, including polyps, cancer, diverticular disease, or colitis. ResultsThe outcome and pathology reports of all patients were obtained and assessed. Clear correlation between actual data value and artificial neural network value were found (r=0.931; p=0.0001). The predictive accuracy of the neural network was 95% in the training group and was 89% (95% CI 84 to 96) in the validation set. This accuracy was significantly higher than the clinical accuracy (69%). ConclusionsArtificial neural networks are more accurate (89% correlation) than standard statistics (67%) when applied to the prediction in individual patients of the need for lower gastrointestinal endoscopy. The results obtained highlight their obvious usefulness, which could now be used in a prospective evaluation for application of the technique.
Gardiner, A., Maslekar, S., & Duthie, G. (2006). Can artificial neural networks predict which patients need a colonoscopy?. Gut : journal of the British Society of Gastroenterology, 55, (A23 - A23). ISSN 0017-5749
|Journal Article Type||Meeting Abstract|
|Acceptance Date||Apr 30, 2006|
|Publication Date||Apr 30, 2006|
|Publisher||BMJ Publishing Group|
|Peer Reviewed||Peer Reviewed|
|Pages||A23 - A23|
|Keywords||artificial neural networks, colonoscopy, prediction|
You might also like
Quality assurance in flexible sigmoidoscopy: Medical and nonmedical endoscopists