I. Tarrio

I. Tarrio

Auto-generated profile, based on publicly available information

I. Tarrio

ULSAM – Hospital de Santa Luzia, Viana do Castelo, Portugal

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
14 April 2023 in Endoscopy
Verified
Journal Article
Closed access
ITI. Tarrio
I. Tarrio
MMM. Moreira
M. Moreira
JRJosé María Ramada Rodilla
José María Ramada Rodilla
TAT. Araújo
T. Araújo
LLLuis Lopes
Luis Lopes
Text An 89-year-old man underwent an ampulectomy of a 40 mm ampullary adenoma. A 15 mm duodenal perforation was identified in the endoscopic scar, in the duodenal wall below the ampulla. The perforation was closed with an OTCS, after placing a fully covered metallic stent in the common bile duct (CBD), to avoid inadvertent closure during the implantation of the OTS-clip. The patient was discharged several weeks after the procedure, and 6 months after the ampulectomy, he remains assymptomatic, with no signs of residual adenoma on follow-up duodenoscopy [1] [2]. Article published online: 14 April 2023 © 2023. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Rüdigerstraße 14, 70469 Stuttgart, Germany

14 April 2023 in Endoscopy
Verified
Journal Article
Closed access
ITI. Tarrio
I. Tarrio
MMM. Moreira
M. Moreira
DFDavid Douglas De Sousa Fernandes
David Douglas De Sousa Fernandes
PSP. F. Silva
P. F. Silva
TAT. Araújo
T. Araújo
LLLuis Lopes
Luis Lopes
Aims Colorectal cancer is the 3rd most prevalent and 2nd deadliest worldwide, so that an efficient screening is essential. The ColonCAD project aims to and Artificial Intelligence based system, that allows the automatic segmentation, detection, and classification (SD&C) of colon polyps. We aimed to assess ColonCAD's accuracy in the SD&C of polyps in colonoscopy images. Methods A convolutional neural network (CNN) algorithm was developed for the SD&C of colon polyps (adenomas or hyperplastic). 3181 images of 644 polyps were used, from a database of patients submitted to polypectomy in a single-center, between november 2019-august 2022. The training data-set comprised 70% of the pooled images, with the remaining being used for validation. The performance of the CNN was compared to delimitation of polyps by 2 colonoscopy experts. The performance of ColonCAD for segmentation was evaluated by Dice coefficient and F1-score. For detection, mean Average Precision and F1-score were used. Sensitivity, specificity and negative/positive predictive values (NPV/PPV) were used to assess classification. Results The segmentation model showed a Dice coefficient of 0.93±0.14 and F1-Score of 0,92±0,33. The detection model achieved a mean Average Precision of 89,3% and F1-Score of 1±0.2. The CNN classification model presented an overall accuracy of 90%. Adenomas were detected with a sensitivity of 80%, 82.4% of specificity, 86.9% of VPP and 73.7% of VPN. For hyperplastic polyps, sensitivity was 82.4%, specificity 80%, VPP 73.7%, and VPN 86.9%. Conclusions ColonCAD platform successfully identified colon polyps and allowed their differentiation into adenoma or hyperplastic with high accuracy. Article published online: 14 April 2023 © 2023. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Rüdigerstraße 14, 70469 Stuttgart, Germany