Sources of performance variability in deep learning-based polyp detection.

PURPOSE: Validation metrics are a key prerequisite for the reliable tracking of scientific progress and for deciding on the potential clinical translation of methods. While recent initiatives aim to develop comprehensive theoretical frameworks for understanding metric-related pitfalls in image analysis problems, there is a lack of experimental evidence on the concrete effects of common and rare pitfalls on specific applications. We address this gap in the literature in the context of colon cancer screening. METHODS: Our contribution is twofold. Firstly, we present the winning solution of the Endoscopy Computer Vision Challenge on colon cancer detection, conducted in conjunction with the IEEE International Symposium on Biomedical Imaging 2022. Secondly, we demonstrate the sensitivity of commonly used metrics to a range of hyperparameters as well as the consequences of poor metric choices. RESULTS: Based on comprehensive validation studies performed with patient data from six clinical centers, we found all commonly applied object detection metrics to be subject to high inter-center variability. Furthermore, our results clearly demonstrate that the adaptation of standard hyperparameters used in the computer vision community does not generally lead to the clinically most plausible results. Finally, we present localization criteria that correspond well to clinical relevance. CONCLUSION: We conclude from our study that (1) performance results in polyp detection are highly sensitive to various design choices, (2) common metric configurations do not reflect the clinical need and rely on suboptimal hyperparameters and (3) comparison of performance across datasets can be largely misleading. Our work could be a first step towards reconsidering common validation strategies in deep learning-based colonoscopy and beyond.

[New methods for endoscopic hemostasis: focus on non-variceal gastrointestinal bleeding]. / Neue Methoden zur endoskopischen Blutstillung: Fokus auf die nicht variköse Blutung.

Gastrointestinal bleeding is a frequent emergency in daily clinical practice of a gastroenterologist. While incidence and mortality of gastrointestinal bleeding are decreasing in many countries, numbers of endoscopic procedures are increasing. Endoscopic therapy of non-variceal gastrointestinal bleeding is still mainly based on "classical" procedures like injection of vasoactive drugs (i. e. epinephrine) or blood derivates, application of through-the-scope hemoclips (TTSC), Argon plasma coagulation and bipolar coagulation. However, in the last years new endoscopic techniques especially for non-variceal gastrointestinal bleedings have become available and enriched our endoscopic equipment. For example, over-the-scope clips (OTSCs) surpass the size of TTSCs and have been successfully established for treatment of gastrointestinal bleeding and leak closure of fistulas and perforations. In addition, hemostatic powders were shown to achieve primary hemostasis in several cases of gastrointestinal bleeding. Besides a brief overview of "classical" endoscopic procedures for hemostasis of non-variceal gastrointestinal bleeding, this review focuses on new epidemiological data and uprising methods for endoscopic hemostasis.

A Simple Evaluation Tool (ET-CET) Indicates Increase of Diagnostic Skills From Small Bowel Capsule Endoscopy Training Courses: A Prospective Observational European Multicenter Study.

Small bowel capsule endoscopy (SBCE) has become a first line diagnostic tool. Several training courses with a similar format have been established in Europe; however, data on learning curve and training in SBCE remain sparse.Between 2008 and 2011, different basic SBCE training courses were organized internationally in UK (n = 2), Italy (n = 2), Germany (n = 2), Finland (n = 1), and nationally in Germany (n = 10), applying similar 8-hour curricula with 50% lectures and 50% hands-on training. The Given PillCam System was used in 12 courses, the Olympus EndoCapsule system in 5, respectively. A simple evaluation tool for capsule endoscopy training (ET-CET) was developed using 10 short SBCE videos including relevant lesions and normal or irrelevant findings. For each video, delegates were required to record a diagnosis (achievable total score from 0 to 10) and the clinical relevance (achievable total score 0 to 10). ET-CET was performed at baseline before the course and repeated, with videos in altered order, after the course.Two hundred ninety-four delegates (79.3% physicians, 16.3% nurses, 4.4% others) were included for baseline analysis, 268 completed the final evaluation. Forty percent had no previous experience in SBCE, 33% had performed 10 or less procedures. Median scores for correct diagnosis improved from 4.0 (IQR 3) to 7.0 (IQR 3) during the courses (P < 0.001, Wilcoxon), and for correct classification of relevance of the lesions from 5.0 (IQR 3) to 7.0 (IQR 3) (P < 0.001), respectively. Improvement was not dependent on experience, profession, SBCE system, or course setting. Previous experience in SBCE was associated with higher baseline scores for correct diagnosis (P < 0.001; Kruskal-Wallis). Additionally, independent nonparametric partial correlation with experience in gastroscopy (rho 0.33) and colonoscopy (rho 0.27) was observed (P < 0.001).A simple ET-CET demonstrated significant improvement of diagnostic skills on completion of formal basic SBCE courses with hands-on training, regardless of preexisting experience, profession, and course setting. Baseline scores for correct diagnoses show a plateau after interpretation of 25 SBCE before courses, supporting this number as a compromise for credentialing. Experience in flexible endoscopy may be useful before attending an SBCE course.

Endoscopic findings in patients with eosinophilic esophagitis.

BACKGROUND: Endoscopy has a key role in establishing the diagnosis of eosinophilic esophagitis (EoE), but endoscopic features of EoE might not be well known. METHODS: All patients aged 18 or older who were diagnosed with EoE from 2008 to 2013 were systematically identified retrospectively and findings at esophago-gastro-duodenoscopy (EGD) were reviewed by two experienced endoscopists through a query of the university hospital database. Patients in whom biopsies from the esophagus were lacking or inadequate for histopathological examination were excluded. RESULTS: 23 patients (17 male, 6 female) were included into the study (median age: 38 years, range: 19 to 71 years). Patients presented with the following symptoms: 12 (52 %) had bolus obstruction and 18 (78 %) dysphagia and/or chest pain. At EGD, 22 of 23 (96 %) patients were observed with at least one endoscopic feature of EoE, i. e., mucosal edema (52 %), longitudinal furrows (57 %), vertical furrows (48 %), or crêpe paper esophagus (52 %). CONCLUSIONS: Typical endoscopic features were present in most patients in whom EoE was diagnosed. Recognizing typical characteristics of EoE is substantial for establishing the diagnosis and for taking biopsies.

Identifying indications for percutaneous (PTC) vs. endoscopic ultrasound (EUS)- guided "rendezvous" procedure in biliary obstruction and incomplete endoscopic retrograde cholangiography (ERC).

BACKGROUND: The variety of rendezvous (RV) procedures has recently been extended by EUS- and PTCD-guided procedures as a complementary means to conventional ERCP. We have identified indication criteria and the potential of biliary PTCD-guided vs. EUS-guided RV. METHODS: Consecutive patients with bile duct obstruction who underwent RV were included. In all, ERCP alone was unable to achieve treatment success. Indication, technical success, and outcome in PTCD- vs. EUS-guided RV were retrospectively compared to identify criteria that indicate preference of RV technique. Site of obstruction, clinical scenario (stenosis with abscess vs. no abscess) and reason for previous failure of ERC were evaluated. RESULTS: In 32 patients, three different indications for RV procedures were identified: First, a one-step access to assist in failed ERCP (type 1, intra-ductal RV); second, temporary drainage for prolonged treatment of complex biliary disease (type 2, intra-ductal RV), and drainage of cholangio-abscess with re-establishing bile outflow (type 3, intra-abscess RV). Indication of PTCD- vs. EUS-guided rendezvous was competitive in type 1, but exclusive in favor of PTCD in types 2 and 3. The site of biliary obstruction indicated the anatomic location of RV procedures. CONCLUSIONS: This classification may help to define inclusion criteria for prospective studies on biliary RV procedures. Choice of therapeutic strategy depends on the anatomic location of the biliary obstruction and the type of the biliary lesion. PTCD-guided RV might improve outcome in cholangio-abscess.

Self-expandable metal stent for malignant colonic obstruction: outcome in proximal vs. left sided tumor localization.

INTRODUCTION: The aim of this study was to evaluate the outcome of through-the-scope (TTS) implanted self-expanding metal stent (SEMS) comparing left-sided vs. proximal placement with regard to complications and outcome in palliation of malignant colorectal obstruction. MATERIAL AND METHODS: All patients were consecutively retrospectively enrolled to this study between January 2009 and February 2012 due to impending or prevalent complete malignant colorectal obstruction. TTS applicable uncovered nitinol SEMS with unique flexible properties were used (Taewoong Medical, South Korea). Left-sided obstruction (aboral from the left flexure) was compared to proximal (from the ileo-cecal valve to the left flexure) localization. All patients have been discussed in the interdisciplinary tumor conference and the recommendation to treat by endoscopic stent placement was given in consensus. RESULTS: A total of 15 patients was enrolled to this study (10 male and 5 female; mean age 68.3 ± 15.4 years, range 48 - 94), five patients with obstructions located in the proximal hemicolon whereas ten patients had a left-sided malignancy. Technical success was achieved in all cases and there was no early complication noticed. Three late complications included tumor overgrowth (n = 1), stent occlusion (1), and dislocation (1). Stent-in-stent insertion achieved, again, clinical success. The site of SEMS implantation (proximal vs. left colon) had no impact on patient outcome or complication rate. SEMS patency duration was 269.8 ± 175.2 days (range 30 - 570) and mean survival of the patients was 305.1 ± 279.3 days (range 16 - 990). CONCLUSION: TTS application of flexible, non-covered SEMS seems to be safe and effective for palliation of malignant colorectal obstruction independent of localization of the tumor in the colon.

[Diagnosis of intestinal metastases from malignant melanoma]. / Diagnostik intestinaler Metastasen bei Malignem Melanom.

The small bowel harbors metastases of malignant melanoma in 5 % to 72 %, dependent on tumor stage and evidence of intestinal blood loss. Capsule endoscopy is sensitive in detecting small bowel metastasis. Computed tomography, magnetic resonance imaging, and PET-CT demonstrate extraintestinal intraabdominal metastases. Melanoma patients with signs of intestinal blood loss should be endoscopically investigated, including small bowel endoscopy. Selected patients in advanced tumor stage should also undergo small bowel endoscopy to plan a treatment strategy. A resection of intestinal metastases can improve the prognosis, if all metastases are removed.

Peroral cholangioscopy for diagnosis and therapy of biliary tract disease using an ultra-slim gastroscope.

High-resolution video cholangioscopy is expected to improve diagnostic validity for diseases of the biliary tract. We report our experience in using an ultra-slim gastroscope for diagnosis and treatment of biliary tract disease. Cholangioscopy was attempted in 25 cases (22 patients) and succeeded in 22 cases (success rate 88%; 19 patients). Cholangiocellular carcinoma (CCC) was diagnosed by cholangioscopy in five of 10 cases (histopathologically confirmed in four), or ruled out in five. Cholangioscopy was used to detect stones in mega-choledochus (n=3), to clarify the postoperative condition of the bile ducts (n=2), to diagnose bile duct varices (n=1), and to release a dislodged self-expanding metal stent (n=1), and others. Argon plasma coagulation was successfully completed in a patient with mucin-producing adenomatosis of the bile ducts. One case of non-fatal air embolism occurred before replacing air with CO2 insufflation. In summary, peroral cholangioscopy with an ultra-slim gastroscope is feasible and helpful in selected patients, improving diagnostic validity, and offering new therapeutic interventions. This technique should only be performed using CO2 insufflation.

Algorithm for detection of small-bowel metastasis in malignant melanoma of the skin.

BACKGROUND AND STUDY AIM: The aim of this study was to develop an algorithm to detect small-bowel metastasis (SBM) of melanoma by sequential laboratory parameters and pan-intestinal endoscopy (PIE) including video capsule endoscopy (VCE). PATIENTS AND METHODS: A total of 390 melanoma patients (AJCC stage I/II/III/IV, 140/80/121/49) were screened for signs of intestinal blood loss (fecal occult blood test [FOBT] or overt bleeding) in an open, multicenter, prospective study, and those who were positive underwent PIE. Independent of the presence of intestinal bleeding, all stage IV patients were offered PIE. Follow-up was obtained in 357 patients (91.5 %) for a median of 16 months. We undertook to identify possible associations between SBM and clinical and laboratory data. Survival data were analyzed with regard to clinical and laboratory data and small-bowel findings. RESULTS: Intestinal blood loss was suspected in 49 of 390 patients (12.6 %), 38 of whom (77.6 %) agreed to undergo endoscopy. In 10 patients, SBM was detected by VCE (intention-to-diagnose, 20.4 %; AJCC III, n = 2; AJCC IV, n = 8). The SBM was resected in five patients. Total detection rates of SBM were 14 of 49 patients in stage IV (28.6 %, intention-to-diagnose), 2 of 121 in stage III (1.7 %), and 0 in stage I/II. In FOBT-positive patients, SBM detection rates were 72.7 %, 14.3 %, and 0 % in tumor stages IV, III, and I/II, respectively. Positive FOBT proved to be an independent negative prognostic factor for total survival in stage III and IV melanoma. CONCLUSIONS: SBMs are frequent in advanced melanoma. In stage III patients, screening for intestinal blood loss by PIE may help to identify SBMs. In stage IV, indication for PIE should depend on the individual consequences of detecting SBM, but not on bleeding symptoms alone.

[Endoscopic-retrograde cholangiopancreatography in patients with surgical modification of anatomy]. / Endoskopisch-retrograde Cholangiopankreatografie (ERCP) bei Patienten mit operativ veränderter Anatomie des Gastrointestinaltrakts.

Endoscopic retrograde cholangiopancreatography (ERCP) is an essential element in treating disease of the bilio-pancreatic system. In some patients, access to the bile ducts is limited due to operatively altered anatomy. The aim of this review is to illustrate the endoscopic procedure in these patients. We present the available evidence and comment on our approach to ERCP in patients with surgical modification of anatomy. In conclusion, conventional side-viewing or forward-viewing endoscopes allow one to access the biliary system in most patients with Billroth-II or partial pancreaticoduodenectomy. By use of balloon enteroscopy, biliary intervention in spite of surgical reconstruction with a long limb - such as Roux-en-Y gastrojejunostomy or hepaticojejunostomy and gastric bypass for obesity - has become feasible in many cases. Adaption of accessory devices to balloon-assisted enteroscopes permits therapeutic interventions in these patients.

[Small bowel diverticula - unknown disease]. / Die Divertikelkrankheit des Dünndarms im Zeitalter der Dünndarmendoskopie.

Diverticula of the small bowel are quite frequent (about 5 %), but being a differential diagnosis of abdominal complaints they may still be underdiagnosed. One reason for the low level of suspicion for a complication of small bowel diverticula might be that the small bowel was out of the gastroenterologists' focus until recently when small bowel endoscopy became available as a reliable and practical diagnostic tool. Diverticula of the jejunum and the ileum may lead to diverticulitis, abscess, obstruction, bleeding, and perforation. Small intestinal bacterial overgrowth syndrome due to small bowel diverticula is a common complication and involves meteorism and malassimilation syndrome. Meckel's diverticulum sometimes provokes bleeding even in the young adult and resection of the diverticulum is clearly indicated. Contrariwise, incidental detection of a Meckel's diverticulum during abdominal laparotomy does not necessarily imply resection of the diverticulum. The juxtapapillary duodenal diverticula appear to be a risk factor for gallbladder stones, bile duct stones, and their recurrence. Moreover, the complication rate of endoscopic interventions of the bile system might be increased. In this review, we highlight the historical background of small bowel diverticular disease, present the epidemiological and pathophysiological implications and provide information on the diagnostic modalities that are available. Diseases of the primary acquired and congential small bowel diverticula are described in detail.

Role of small-bowel endoscopy in the management of patients with inflammatory bowel disease: an international OMED-ECCO consensus.

Crohn's disease and ulcerative colitis are lifelong diseases seen predominantly in the developed countries of the world. Whereas ulcerative colitis is a chronic inflammatory condition causing diffuse and continuous mucosal inflammation of the colon, Crohn's disease is a heterogeneous entity comprised of several different phenotypes, but can affect the entire gastrointestinal tract. A change in diagnosis from Crohn's disease to ulcerative colitis during the first year of illness occurs in about 10 % - 15 % of cases. Inflammatory bowel disease (IBD) restricted to the colon that cannot be characterized as either ulcerative colitis or Crohn's disease is termed IBD-unclassified (IBDU). The advent of capsule and both single- and double-balloon-assisted enteroscopy is revolutionizing small-bowel imaging and has major implications for diagnosis, classification, therapeutic decision making and outcomes in the management of IBD. The role of these investigations in the diagnosis and management of IBD, however, is unclear. This document sets out the current Consensus reached by a group of international experts in the fields of endoscopy and IBD at a meeting held in Brussels, 12-13th December 2008, organised jointly by the European Crohn's and Colitis Organisation (ECCO) and the Organisation Mondiale d'Endoscopie Digestive (OMED). The Consensus is grouped into seven sections: definitions and diagnosis; suspected Crohn's disease; established Crohn's disease; IBDU; ulcerative colitis (including ileal pouch-anal anastomosis [IPAA]); paediatric practice; and complications and unresolved questions. Consensus guideline statements are followed by comments on the evidence and opinion. Statements are intended to be read in context with qualifying comments and not read in isolation.

[Small bowel endoscopy]. / Endoskopie des Dünndarms.

Routine endoscopy of the small bowel has found its way into clinical practice within the last few years. Capsule endoscopy (CE) enables with high accuracy detection of the source of mid intestinal bleeding. Crohn's disease of the small bowel is sensitively detected and surveillance in polyposis syndromes is improved by applying sensitive, non-invasive endoscopic means. Double-balloon enteroscopy (DBE) may be used complementary to CE by offering therapeutic possibilities and providing histopathological specimen. This review describes the diagnostic and therapeutic repertoire that is offered by modern small bowel endoscopy.

Course of Crohn's disease prior to establishment of the diagnosis.

BACKGROUND: The course of Crohn's disease prior to the establishment of the diagnosis is widely unknown. Therefore, we instigated a survey amongst newly diagnosed patients. PATIENTS AND METHODS: Patients diagnosed with CD less than 12 months before enrollment were included. Data on demography, social status, time interval to diagnosis, symptoms, and health care service use were collected in a retrospective, web-based, census. Patients were contacted in cooperation with two organizations: a German patients' organization (Deutsche Morbus Crohn/Colitis ulcerosa Vereinigung e.V. [DCCV]) and a professional organization of German gastroenterologists (Berufsverband der Niedergelassenen Gastroenterologen Deutschlands e.V. [bng]). Study participation was anonymous by use of a transaction number. RESULTS: The median interval period between onset of first symptoms and diagnosis was 13 months. During this time, participants reported having five doctor consultations on average, with 44% of them having a mean of 1.5 hospitalizations. 65% were unfit for work with a 14 day median (2 to 480 days) due to their symptoms. A mean (+/-SD) of 8.6 (+/-7.1) diagnostic tests were performed before the diagnosis was established. Overall health state was judged as temporarily bad or very bad by 84% of the participants. Age at diagnosis, characteristic symptoms, and localization of the disease for the participants did not differ from previously reported international data. DISCUSSION: This web-based survey shows a substantial time interval of over one year until diagnosis of Crohn's disease amongst the study participants. This period is characterized by both psychological stress and impaired ability to work.

Diagnosis of small bowel Crohn's disease: a prospective comparison of capsule endoscopy with magnetic resonance imaging and fluoroscopic enteroclysis.

BACKGROUND AND AIMS: The diagnostic yield of capsule endoscopy (CE) compared with magnetic resonance imaging (MRI) in small bowel Crohn's disease is not well established. We prospectively investigated CE, MRI, and double contrast fluoroscopy in patients with suspected small bowel Crohn's disease. METHODS: Fifty two consecutive patients (39 females, 13 males) were investigated by MRI, fluoroscopy and--if bowel obstruction could be excluded--by CE. In 25, Crohn's disease was newly suspected while the diagnosis of Crohn's disease (non-small bowel) had been previously established in 27. RESULTS: Small bowel Crohn's disease was diagnosed in 41 of 52 patients (79%). CE was not accomplished in 14 patients due to bowel strictures. Of the remaining 27 patients, CE, MRI, and fluoroscopy detected small bowel Crohn's disease in 25 (93%), 21 (78%), and 7 (of 21; 33%) cases, respectively. CE was the only diagnostic tool in four patients. CE was slightly more sensitive than MRI (12 v 10 of 13 in suspected Crohn's disease and 13 v 11 of 14 in established Crohn's disease). MRI detected inflammatory conglomerates and enteric fistulae in three and two cases, respectively. CONCLUSION: CE and MRI are complementary methods for diagnosing small bowel Crohn's disease. CE is capable of detecting limited mucosal lesions that may be missed by MRI, but awareness of bowel obstruction is mandatory. In contrast, MRI is helpful in identifying transmural Crohn's disease and extraluminal lesions, and may exclude strictures.

ERCP and MRCP--when and why.

Since the introduction of endoscopic retrograde cholangiopancreatography (ERCP) in the 1970s, gastroenterologists have a wide spectrum of diagnostic and therapeutic options in the biliopancreatic ductal system at their disposal. With its arrival in the 1990s, magnetic resonance cholangiopancreatography (MRCP) developed as a potent diagnostic tool in biliopancreatic pathology. Currently, MRCP is widely replacing diagnostic ERCP and thereby avoiding complications related to endoscopic technique.We summarize evidence-based data and demonstrate indications and differential indications for MRCP and ERCP in pancreatic disease. Complications related to the procedures and possible medical prevention are discussed. The feasibility of interventional endoscopy in pancreatic disease is reported in detail. The role of gastroenterologists in performing MRCP is outlined on the basis of practical examples.