Endoscopy 2018; 50(04): S148-S149
DOI: 10.1055/s-0038-1637479
ESGE Days 2018 ePosters
Georg Thieme Verlag KG Stuttgart · New York

DEVELOPMENT OF ALGORITHM FOR ASSESSMENT OF NEOPLASTIC LESIONS OF THE SMALL BOWEL ON THE BASIS OF VIDEO CAPSULE ENDOSCOPY TO SUPPORT CLINICAL DECISION

DE Seleznev
1   N.I.Pirogov Russian National Research Medical University. Medical Rehabilitation Center “Klinika+31”. Moscow University Hospital №31, Moscow, Russian Federation
,
AV Budykina
2   N.I. Pirogov Russian National Research Medical University, Moscow, Russian Federation
,
SE Rauzina
2   N.I. Pirogov Russian National Research Medical University, Moscow, Russian Federation
,
ED Fedorov
3   Moscow University Hospital №31, Department of Abdominal Surgery, Surgical Gastroenterology and Endoscopy, Moscow, Russian Federation
,
EV Ivanova
4   Rehabilitation Center “Klinika+31”, Moscow, Russian Federation
› Author Affiliations
Further Information

Publication History

Publication Date:
27 March 2018 (online)

Also available ateRef
 

Aims:

Even for experienced decryptor search of tumors in video capsule record sometime presents significant difficulties. We made an attempt to develop rules for automated processing of endoscopic images, obtained during video capsule endoscopy (VCE).

Methods:

Retrospectively we've carefully studied video sequences and snapshort images of 181 neoplastic lesions of jejunum and ileum, obtained by VCE from 65 patients (m-35, f-30, mean age 46 ± 28yrs., range 18 – 80). Each neoplastic lesion was histologically verified after it's endoscopic or surgical removal. According to expert's opinion we initially created a list of 30 features and their gradations, which were important for the assessment of neoplastic diseases of the jejunum and ileum on the VCE images.

Results:

Of the 30 selected features eight characteristics (gender of a patient; deformation of the wall/lumen of the intestine; change of the small bowel folds’ direction; polypoid changes of the mucosa; changes in vascular mucosal pattern; mucosal irregularity; color changes of mucosa and lobed structure of neoplasia) were statistically significant, influencing the division of the studied objects into groups. Using heterogeneous Bayesian diagnostic procedures and the calculation of the diagnostic ratios three-level algorithm for differential diagnosis of neoplastic diseases of the jejunum and ileum was developed. We've got a satisfactory distribution of research objects into 4 groups: non-neoplastic lesions of the small bowel (sensitivity = 86%, specificity = 92%); epithelial benign tumors of the small intestine (sensitivity = 89%, specificity = 93%); non-epithelial benign tumors of the small intestine (sensitivity = 86%, specificity = 97%); malignant tumors of the small intestine (sensitivity = 89%, specificity = 93%). The elaborated algorithm was implemented as a software module “VCE conclusion” with integrated development environment Visual Studio and the programming language C#.

Conclusions:

Algorithm for assessment of neoplastic lesions of the small bowel during VCE is promising trend. It could become valuable tool for support clinical decision of a doctor analyzing video capsule endoscopy recording.