- Email: [email protected]
Su1661 Risk Factors for Technical Failure of Endoscopic Multiple Self-Expandable Metallic Stent Placement by Stent-in-Stent Method
Abstracts
INDICATION 1) Biliary Leak a) Post-cholecystectomy b) Abscess/Pancreatic necrosis c) Post-EUS-Biliary rendezvous 2) Bleeding a) Post-sphincterotomy b) Bile duct varices c) Bleeding Neuro-endocrine tumor 3) Perforation a) Post-sphincterotomy b) Biliary-enteric anastomotic 4) Other a) Removal of partially covered self-expandable stent b) Facilitate biliary stone removal c) Sump Syndrome d) Choledochogastric fistula
35 24 10 1 27 23 3 1 18 12 6 7 3 2 1 1
Post-procedural details of patients undergoing covered selfexpandable metal stent placement in non-stricture related benign biliary diseases (n[87) Prior plastic stent use, n (%) Type of stent, n (%) 1) Wallstent 2) Wallflex (10mm) 3) Viabil Median duration of stent placement (weeks) 1) Biliary Leak 2) Bleeding 3) Perforation Biliary leak resolved, n (%) Bleeding resolved, n (%) Perforation resolved, n (%) a) Removal of partially covered self-expandable metal stent b) Facilitate biliary stone removal c) Sump Syndrome d) Choledochogastric fistula
32 (37%) 87 12 (14%) 29 (33%) 46 (53%) 9 (2 - 115) 3 (1 - 11) 9.5 (4- 14) 33/35 (94%) * 25/27 (93%) ** 18/18 (100%) 2/3 (67%) 1/2 (50%) 0/1 (0%) 0/1 (0%)
* 1 patient died from sepsis from infected biloma and another patient failed. ** 1 patient rebled after being restarted on Coumadin, and another while on clopidogrel.
Su1662 Smart Atlas for Supporting the Interpretation of Probe-Based Confocal LASER Endomicroscopy (pCLE) of Biliary Strictures: First Classification Results of a Computer-Aided Diagnosis Software Based on Image Recognition Marzieh Kohandani Tafreshi*1,2, Virendra Joshi3, Alexander Meining4, Charles J. Lightdale5, Marc Giovannini6, Julien Dauguet2, Nicholas Ayache1, Barbara André2 1 INRIA, Sophia Antipolis, France; 2Mauna Kea Technologies, Paris, France; 3Ochsner Medical Center Kenner, Kenner, LA; 4Klinikum rechts der Isar, München, Germany; 5Columbia University Medical Center, New York, NY; 6Institut Paoli-Calmettes, Marseille, France Background and Aims: pCLE enables microscopic imaging of biliary strictures, in vivo and in real time, during an ERCP procedure. Results of a multicentric study (Meining et al., GIE 2011) have shown that pCLE allows endoscopists to differentiate benign from malignant strictures in real time with high sensitivity and NPV. A computer-aided diagnosis software called Smart Atlas has been developed to assist endoscopists with the interpretation of pCLE sequences. This study aims at evaluating the performance of this software for the differentiation of benign and malignant strictures. Methods: Several high quality pCLE sequences were retrospectively collected from pCLE procedures performed in multiple clinical centers. These sequences, along with their annotated final diagnosis, were used to train a classification software that uses a content-based image retrieval algorithm to predict the diagnosis of a query video based on the diagnoses of the most visually similar atlas videos. For all cases, final diagnosis was based on histology, positive tissue sampling, or one year follow-up. All evaluations were performed using leave-one-patient-out cross-validation to avoid bias. To evaluate binary classification, a receiver operating curve was generated, allowing optimization of the trade-off between false positives and negatives. Results: Among the 60 pCLE sequences collected from 30 patients, 14 were representative of healthy bile duct, 10 of inflammatory strictures and 36 of malignant strictures. The resulting receiver operating curve shows two points of interest: the first (reps. second) point has a high sensitivity of 88.9% (reps. high specificity of 91.7%), an acceptable specificity of 70.8% (reps. acceptable sensitivity of 69.4%), an accuracy of 81.7% (resp. 78.3%), a PPV of 82.1% (resp. 92.6%) and a NPV of 81.0% (resp. 66.7%). In comparison, Meining et al. reported that, for in vivo pCLE diagnosis of malignant stricture, endoscopists achieve overall sensitivity, specificity, accuracy, PPV and NPV of 98%, 67%, 81%, 71% and 97%, respectively. Limitations: Small and unbalanced sample size, restricted to high quality videos. Conclusions: These first results demonstrate that benign and malignant strictures can be automatically discriminated by the Smart Atlas software using only the image content of pCLE sequences of high quality, with an accuracy comparable to that achieved in real-time by endoscopists. The software is also able to achieve high specificity and PPV to help reduce false positives caused by inflammatory strictures. Future work will focus on improving the software to handle pCLE sequences of various quality. The resulting case-based reasoning software could be used as an educational tool to train non-expert endoscopists, but also as a second-reader tool to assist any endoscopist in real-time diagnosis of biliary strictures using pCLE.
Su1661 Risk Factors for Technical Failure of Endoscopic Multiple SelfExpandable Metallic Stent Placement by Stent-in-Stent Method Kazumichi Kawakubo*, Hiroshi Kawakami, Masaki Kuwatani, Yoko Abe, Taiki Kudo, Kimitoshi Kubo, Yoshimasa Kubota, Naoya Sakamoto Department of Gastroenterology and Hepatology, Hokkaido University Graduate School of Medicine, Sapporo, Japan Background: Endoscopic multiple self-expandable metallic stent (SEMS) placement by the stent-in-stent (SIS) method has been reported to be useful for the management of unresectable hilar malignant biliary obstruction. However, it is technically challenging, and the optimal SEMS for the procedure has remained unknown. The aim of this study was to identify the risk factors for technical failure of endoscopic multiple SEMS placement for unresectable malignant hilar biliary obstruction (MHBO). Methods: Between December 2009 and May 2013, 50 consecutive patients with MHBO underwent endoscopic multiple SEMS placement by the SIS method. We retrospectively evaluated the rate of successful multiple SEMS placement and identified the risk factors for technical failure. Results: The technical success rate for multiple SEMS placement was 82.0% (95% confidence interval [CI]: 69.2-90.2). On univariate analysis, the rate of technical failure was high in patients with metastatic disease and unilateral placement. Multivariate analysis revealed that metastatic disease was a significant risk factor for technical failure (odds ratio: 9.63, 95% CI: 1.11105.5). Subgroup analysis after multiple guidewire insertion showed that the rate of technical success was higher in the laser-cut than in the braided type SEMS. Conclusions: Metastatic disease was a significant risk factor for technical failure of multiple SEMS placement for unresectable MHBO. A laser-cut SEMS might be preferable for the SIS procedure. (UMIN-CTR; No. UMIN000011879)
Receiver operating characteristic curve for the binary classification between benign and malignant biliary strictures.
AB250 GASTROINTESTINAL ENDOSCOPY Volume 79, No. 5S : 2014
www.giejournal.org
INDICATION 1) Biliary Leak a) Post-cholecystectomy b) Abscess/Pancreatic necrosis c) Post-EUS-Biliary rendezvous 2) Bleeding a) Post-sphincterotomy b) Bile duct varices c) Bleeding Neuro-endocrine tumor 3) Perforation a) Post-sphincterotomy b) Biliary-enteric anastomotic 4) Other a) Removal of partially covered self-expandable stent b) Facilitate biliary stone removal c) Sump Syndrome d) Choledochogastric fistula
35 24 10 1 27 23 3 1 18 12 6 7 3 2 1 1
Post-procedural details of patients undergoing covered selfexpandable metal stent placement in non-stricture related benign biliary diseases (n[87) Prior plastic stent use, n (%) Type of stent, n (%) 1) Wallstent 2) Wallflex (10mm) 3) Viabil Median duration of stent placement (weeks) 1) Biliary Leak 2) Bleeding 3) Perforation Biliary leak resolved, n (%) Bleeding resolved, n (%) Perforation resolved, n (%) a) Removal of partially covered self-expandable metal stent b) Facilitate biliary stone removal c) Sump Syndrome d) Choledochogastric fistula
32 (37%) 87 12 (14%) 29 (33%) 46 (53%) 9 (2 - 115) 3 (1 - 11) 9.5 (4- 14) 33/35 (94%) * 25/27 (93%) ** 18/18 (100%) 2/3 (67%) 1/2 (50%) 0/1 (0%) 0/1 (0%)
* 1 patient died from sepsis from infected biloma and another patient failed. ** 1 patient rebled after being restarted on Coumadin, and another while on clopidogrel.
Su1662 Smart Atlas for Supporting the Interpretation of Probe-Based Confocal LASER Endomicroscopy (pCLE) of Biliary Strictures: First Classification Results of a Computer-Aided Diagnosis Software Based on Image Recognition Marzieh Kohandani Tafreshi*1,2, Virendra Joshi3, Alexander Meining4, Charles J. Lightdale5, Marc Giovannini6, Julien Dauguet2, Nicholas Ayache1, Barbara André2 1 INRIA, Sophia Antipolis, France; 2Mauna Kea Technologies, Paris, France; 3Ochsner Medical Center Kenner, Kenner, LA; 4Klinikum rechts der Isar, München, Germany; 5Columbia University Medical Center, New York, NY; 6Institut Paoli-Calmettes, Marseille, France Background and Aims: pCLE enables microscopic imaging of biliary strictures, in vivo and in real time, during an ERCP procedure. Results of a multicentric study (Meining et al., GIE 2011) have shown that pCLE allows endoscopists to differentiate benign from malignant strictures in real time with high sensitivity and NPV. A computer-aided diagnosis software called Smart Atlas has been developed to assist endoscopists with the interpretation of pCLE sequences. This study aims at evaluating the performance of this software for the differentiation of benign and malignant strictures. Methods: Several high quality pCLE sequences were retrospectively collected from pCLE procedures performed in multiple clinical centers. These sequences, along with their annotated final diagnosis, were used to train a classification software that uses a content-based image retrieval algorithm to predict the diagnosis of a query video based on the diagnoses of the most visually similar atlas videos. For all cases, final diagnosis was based on histology, positive tissue sampling, or one year follow-up. All evaluations were performed using leave-one-patient-out cross-validation to avoid bias. To evaluate binary classification, a receiver operating curve was generated, allowing optimization of the trade-off between false positives and negatives. Results: Among the 60 pCLE sequences collected from 30 patients, 14 were representative of healthy bile duct, 10 of inflammatory strictures and 36 of malignant strictures. The resulting receiver operating curve shows two points of interest: the first (reps. second) point has a high sensitivity of 88.9% (reps. high specificity of 91.7%), an acceptable specificity of 70.8% (reps. acceptable sensitivity of 69.4%), an accuracy of 81.7% (resp. 78.3%), a PPV of 82.1% (resp. 92.6%) and a NPV of 81.0% (resp. 66.7%). In comparison, Meining et al. reported that, for in vivo pCLE diagnosis of malignant stricture, endoscopists achieve overall sensitivity, specificity, accuracy, PPV and NPV of 98%, 67%, 81%, 71% and 97%, respectively. Limitations: Small and unbalanced sample size, restricted to high quality videos. Conclusions: These first results demonstrate that benign and malignant strictures can be automatically discriminated by the Smart Atlas software using only the image content of pCLE sequences of high quality, with an accuracy comparable to that achieved in real-time by endoscopists. The software is also able to achieve high specificity and PPV to help reduce false positives caused by inflammatory strictures. Future work will focus on improving the software to handle pCLE sequences of various quality. The resulting case-based reasoning software could be used as an educational tool to train non-expert endoscopists, but also as a second-reader tool to assist any endoscopist in real-time diagnosis of biliary strictures using pCLE.
Su1661 Risk Factors for Technical Failure of Endoscopic Multiple SelfExpandable Metallic Stent Placement by Stent-in-Stent Method Kazumichi Kawakubo*, Hiroshi Kawakami, Masaki Kuwatani, Yoko Abe, Taiki Kudo, Kimitoshi Kubo, Yoshimasa Kubota, Naoya Sakamoto Department of Gastroenterology and Hepatology, Hokkaido University Graduate School of Medicine, Sapporo, Japan Background: Endoscopic multiple self-expandable metallic stent (SEMS) placement by the stent-in-stent (SIS) method has been reported to be useful for the management of unresectable hilar malignant biliary obstruction. However, it is technically challenging, and the optimal SEMS for the procedure has remained unknown. The aim of this study was to identify the risk factors for technical failure of endoscopic multiple SEMS placement for unresectable malignant hilar biliary obstruction (MHBO). Methods: Between December 2009 and May 2013, 50 consecutive patients with MHBO underwent endoscopic multiple SEMS placement by the SIS method. We retrospectively evaluated the rate of successful multiple SEMS placement and identified the risk factors for technical failure. Results: The technical success rate for multiple SEMS placement was 82.0% (95% confidence interval [CI]: 69.2-90.2). On univariate analysis, the rate of technical failure was high in patients with metastatic disease and unilateral placement. Multivariate analysis revealed that metastatic disease was a significant risk factor for technical failure (odds ratio: 9.63, 95% CI: 1.11105.5). Subgroup analysis after multiple guidewire insertion showed that the rate of technical success was higher in the laser-cut than in the braided type SEMS. Conclusions: Metastatic disease was a significant risk factor for technical failure of multiple SEMS placement for unresectable MHBO. A laser-cut SEMS might be preferable for the SIS procedure. (UMIN-CTR; No. UMIN000011879)
Receiver operating characteristic curve for the binary classification between benign and malignant biliary strictures.
AB250 GASTROINTESTINAL ENDOSCOPY Volume 79, No. 5S : 2014
www.giejournal.org