Software as a Medical Device (SaMD) in Digestive Healthcare: Regulatory Challenges and Ethical Implications

The growing integration of software in healthcare, particularly the rise of standalone software as a medical device (SaMD), is transforming digestive medicine, a field heavily reliant on medical imaging for both diagnosis and therapeutic interventions. This narrative review aims to explore the impact of SaMD on digestive healthcare, focusing on the evolution of these tools and their regulatory and ethical challenges. Our analysis highlights the exponential growth of SaMD in digestive healthcare, driven by the need for precise diagnostic tools and personalized treatment strategies. This rapid advancement, however, necessitates the parallel development of a robust regulatory framework to ensure SaMDs are transparent and deliver universal clinical benefits without the introduction of bias or harm. In addition, the discussion highlights the importance of adherence to the FAIR principles for data management-findability, accessibility, interoperability, and reusability. However, enhanced accessibility and interoperability require rigorous protocols to ensure compliance with data protection guidelines and adequate data security, both of which are crucial for effective integration of SaMDs into clinical workflows. In conclusion, while SaMDs hold significant promise for improving patients' outcomes in digestive medicine, their successful integration into clinical workflow depends on rigorous data protection protocols and clinical validation. Future directions include the need for adequate clinical and real-world studies to demonstrate that these devices are safe and well-suited to healthcare settings.

Higher frequency of gastric neoplasia in advanced chronic liver disease patients: Impact of screening endoscopy in an intermediate-high risk country

BACKGROUND

The Baveno VII guidelines were proposed to identify which patients could safely avoid screening esophagogastroduodenoscopy (EGD) for gastroesophageal varices. We aimed to evaluate the frequency of gastric neoplasia in compensated advanced chronic liver disease (cACLD) patients who underwent EGD for screening of gastroesophageal varices (GOEV) compared to a healthy population.

METHODS

Retrospective study that enrolled all cACLD patients who underwent EGD for GOEV screening (January 2008-June 2018) in a tertiary reference center. cACLD patients were compared with asymptomatic healthy individuals who underwent EGD in a private hospital setting (April 2017-March 2018).

RESULTS

We evaluated 1845 patients (481 cACLD patients, 1364 healthy individuals). A significantly higher frequency of gastric neoplasia was observed in patients with cACLD compared to healthy individuals (4.0% vs. 1.0 %; p < 0.001). Rare histopathological subtypes (WHO Classification) accounted for 28.7 % of gastric carcinoma cases in the cACLD cohort. Seven cases of gastric neoplasia (36.8 % of gastric neoplasia cases in the cACLD patients) were diagnosed in patients who, according to the Baveno VII criteria, would have not been submitted to EGD.

CONCLUSION

We found an increased frequency of gastric neoplasia in patients with cACLD in comparison with healthy individuals. In countries with intermediate-high risk for GC, continuing to perform EGD could be beneficial.

Deep learning and minimally invasive inflammatory activity assessment: a proof-of-concept study for development and score correlation of a panendoscopy convolutional network

Background

Capsule endoscopy (CE) is a valuable tool for assessing inflammation in patients with Crohn's disease (CD). The current standard for evaluating inflammation are validated scores (and clinical laboratory values) like Lewis score (LS), Capsule Endoscopy Crohn's Disease Activity Index (CECDAI), and ELIAKIM. Recent advances in artificial intelligence (AI) have made it possible to automatically select the most relevant frames in CE.

Objectives

In this proof-of-concept study, our objective was to develop an automated scoring system using CE images to objectively grade inflammation.

Design

Pan-enteric CE videos (PillCam Crohn's) performed in CD patients between 09/2020 and 01/2023 were retrospectively reviewed and LS, CECDAI, and ELIAKIM scores were calculated.

Methods

We developed a convolutional neural network-based automated score consisting of the percentage of positive frames selected by the algorithm (for small bowel and colon separately). We correlated clinical data and the validated scores with the artificial intelligence-generated score (AIS).

Results

A total of 61 patients were included. The median LS was 225 (0-6006), CECDAI was 6 (0-33), ELIAKIM was 4 (0-38), and SB_AIS was 0.5659 (0-29.45). We found a strong correlation between SB_AIS and LS, CECDAI, and ELIAKIM scores (Spearman's r = 0.751, r = 0.707, r = 0.655, p = 0.001). We found a strong correlation between LS and ELIAKIM (r = 0.768, p = 0.001) and a very strong correlation between CECDAI and LS (r = 0.854, p = 0.001) and CECDAI and ELIAKIM scores (r = 0.827, p = 0.001).

Conclusion

Our study showed that the AI-generated score had a strong correlation with validated scores indicating that it could serve as an objective and efficient method for evaluating inflammation in CD patients. As a preliminary study, our findings provide a promising basis for future refining of a CE score that may accurately correlate with prognostic factors and aid in the management and treatment of CD patients.

Artificial Intelligence and Panendoscopy-Automatic Detection of Clinically Relevant Lesions in Multibrand Device-Assisted Enteroscopy

Device-assisted enteroscopy (DAE) is capable of evaluating the entire gastrointestinal tract, identifying multiple lesions. Nevertheless, DAE's diagnostic yield is suboptimal. Convolutional neural networks (CNN) are multi-layer architecture artificial intelligence models suitable for image analysis, but there is a lack of studies about their application in DAE. Our group aimed to develop a multidevice CNN for panendoscopic detection of clinically relevant lesions during DAE. In total, 338 exams performed in two specialized centers were retrospectively evaluated, with 152 single-balloon enteroscopies (Fujifilm®, Porto, Portugal), 172 double-balloon enteroscopies (Olympus®, Porto, Portugal) and 14 motorized spiral enteroscopies (Olympus®, Porto, Portugal); then, 40,655 images were divided in a training dataset (90% of the images, n = 36,599) and testing dataset (10% of the images, n = 4066) used to evaluate the model. The CNN's output was compared to an expert consensus classification. The model was evaluated by its sensitivity, specificity, positive (PPV) and negative predictive values (NPV), accuracy and area under the precision recall curve (AUC-PR). The CNN had an 88.9% sensitivity, 98.9% specificity, 95.8% PPV, 97.1% NPV, 96.8% accuracy and an AUC-PR of 0.97. Our group developed the first multidevice CNN for panendoscopic detection of clinically relevant lesions during DAE. The development of accurate deep learning models is of utmost importance for increasing the diagnostic yield of DAE-based panendoscopy.

Deep Learning and Minimally Invasive Endoscopy: Automatic Classification of Pleomorphic Gastric Lesions in Capsule Endoscopy

INTRODUCTION

Capsule endoscopy (CE) is a minimally invasive examination for evaluating the gastrointestinal tract. However, its diagnostic yield for detecting gastric lesions is suboptimal. Convolutional neural networks (CNNs) are artificial intelligence models with great performance for image analysis. Nonetheless, their role in gastric evaluation by wireless CE (WCE) has not been explored.

METHODS

Our group developed a CNN-based algorithm for the automatic classification of pleomorphic gastric lesions, including vascular lesions (angiectasia, varices, and red spots), protruding lesions, ulcers, and erosions. A total of 12,918 gastric images from 3 different CE devices (PillCam Crohn's; PillCam SB3; OMOM HD CE system) were used from the construction of the CNN: 1,407 from protruding lesions; 994 from ulcers and erosions; 822 from vascular lesions; and 2,851 from hematic residues and the remaining images from normal mucosa. The images were divided into a training (split for three-fold cross-validation) and validation data set. The model's output was compared with a consensus classification by 2 WCE-experienced gastroenterologists. The network's performance was evaluated by its sensitivity, specificity, accuracy, positive predictive value and negative predictive value, and area under the precision-recall curve.

RESULTS

The trained CNN had a 97.4% sensitivity; 95.9% specificity; and positive predictive value and negative predictive value of 95.0% and 97.8%, respectively, for gastric lesions, with 96.6% overall accuracy. The CNN had an image processing time of 115 images per second.

DISCUSSION

Our group developed, for the first time, a CNN capable of automatically detecting pleomorphic gastric lesions in both small bowel and colon CE devices.

Artificial Intelligence and Anorectal Manometry: Automatic Detection and Differentiation of Anorectal Motility Patterns-A Proof-of-Concept Study

INTRODUCTION

Anorectal manometry (ARM) is the gold standard for the evaluation of anorectal functional disorders, prevalent in the population. Nevertheless, the accessibility to this examination is limited, and the complexity of data analysis and report is a significant drawback. This pilot study aimed to develop and validate an artificial intelligence model to automatically differentiate motility patterns of fecal incontinence (FI) from obstructed defecation (OD) using ARM data.

METHODS

We developed and tested multiple machine learning algorithms for the automatic interpretation of ARM data. Four models were tested: k-nearest neighbors, support vector machines, random forests, and gradient boosting (xGB). These models were trained using a stratified 5-fold strategy. Their performance was assessed after fine-tuning of each model's hyperparameters, using 90% of data for training and 10% of data for testing.

RESULTS

A total of 827 ARM examinations were used in this study. After fine-tuning, the xGB model presented an overall accuracy (84.6% ± 2.9%), similar to that of random forests (82.7% ± 4.8%) and support vector machines (81.0% ± 8.0%) and higher that of k-nearest neighbors (74.4% ± 3.8%). The xGB models showed the highest discriminating performance between OD and FI, with an area under the curve of 0.939.

DISCUSSION

The tested machine learning algorithms, particularly the xGB model, accurately differentiated between FI and OD manometric patterns. Subsequent development of these tools may optimize the access to ARM studies, which may have a significant impact on the management of patients with anorectal functional diseases.

Clinicopathological characteristics and prognostic factors of small bowel lymphomas: a retrospective single-center study

Background

There is little information on diagnosis and management of small bowel lymphomas, and optimal management strategies are still undefined. This study aims to describe their main clinical and pathological characteristics and identify poor prognostic factors.

Methods

A retrospective observational study of all patients with histological diagnosis of small bowel lymphoma between January 2010 and December 2020 was performed.

Results

We included 40 patients, with male predominance (60%) and mean age of 60.7 years. The ileum was the most common location, and the most common histological subtypes were follicular lymphoma and diffuse large B-cell lymphoma. Clinical presentation was variable from asymptomatic patients (30%) to acute surgical complications (35%) including perforation, intestinal obstruction, ileal intussusception, or severe bleeding. Diagnosis was established by endoscopy in 22 patients (55%), and the most common findings included polyps, single mass, diffuse infiltration, or ulceration, whereas 18 (45%) required surgery because of acute presentations or tumor resection, and lymphoma was diagnosed postoperatively. Surgery was curative in one-third of those patients. Median survival was 52 months. Acute presentation (P = 0.001), symptomatic disease (P = 0.003), advanced stage (P = 0.008), diffuse large B-cell lymphoma (P = 0.007), anemia (P = 0.006), hypoalbuminemia (P < 0.001), elevated lactate dehydrogenase (P = 0.02), elevated C-reactive protein (P < 0.001), and absence of treatment response (P < 0.001) were significant predictors of mortality.

Conclusion

Small bowel lymphoma is a rare malignancy with diverse clinical and endoscopic presentations that require a high index of suspicion. Primary factors associated with worse outcome included acute presentation, advanced stage, histological subtype, biochemical abnormalities, and absence of treatment response.

Design of a Convolutional Neural Network as a Deep Learning Tool for the Automatic Classification of Small-Bowel Cleansing in Capsule Endoscopy

Background and objectives: Capsule endoscopy (CE) is a non-invasive method to inspect the small bowel that, like other enteroscopy methods, requires adequate small-bowel cleansing to obtain conclusive results. Artificial intelligence (AI) algorithms have been seen to offer important benefits in the field of medical imaging over recent years, particularly through the adaptation of convolutional neural networks (CNNs) to achieve more efficient image analysis. Here, we aimed to develop a deep learning model that uses a CNN to automatically classify the quality of intestinal preparation in CE. Methods: A CNN was designed based on 12,950 CE images obtained at two clinical centers in Porto (Portugal). The quality of the intestinal preparation was classified for each image as: excellent, ≥90% of the image surface with visible mucosa; satisfactory, 50-90% of the mucosa visible; and unsatisfactory, <50% of the mucosa visible. The total set of images was divided in an 80:20 ratio to establish training and validation datasets, respectively. The CNN prediction was compared with the classification established by consensus of a group of three experts in CE, currently considered the gold standard to evaluate cleanliness. Subsequently, how the CNN performed in diagnostic terms was evaluated using an independent validation dataset. Results: Among the images obtained, 3633 were designated as unsatisfactory preparation, 6005 satisfactory preparation, and 3312 with excellent preparation. When differentiating the classes of small-bowel preparation, the algorithm developed here achieved an overall accuracy of 92.1%, with a sensitivity of 88.4%, a specificity of 93.6%, a positive predictive value of 88.5%, and a negative predictive value of 93.4%. The area under the curve for the detection of excellent, satisfactory, and unsatisfactory classes was 0.98, 0.95, and 0.99, respectively. Conclusions: A CNN-based tool was developed to automatically classify small-bowel preparation for CE, and it was seen to accurately classify intestinal preparation for CE. The development of such a system could enhance the reproducibility of the scales used for such purposes.

The Promise of Artificial Intelligence in Digestive Healthcare and the Bioethics Challenges It Presents

With modern society well entrenched in the digital area, the use of Artificial Intelligence (AI) to extract useful information from big data has become more commonplace in our daily lives than we perhaps realize. Medical specialties that rely heavily on imaging techniques have become a strong focus for the incorporation of AI tools to aid disease diagnosis and monitoring, yet AI-based tools that can be employed in the clinic are only now beginning to become a reality. However, the potential introduction of these applications raises a number of ethical issues that must be addressed before they can be implemented, among the most important of which are issues related to privacy, data protection, data bias, explainability and responsibility. In this short review, we aim to highlight some of the most important bioethical issues that will have to be addressed if AI solutions are to be successfully incorporated into healthcare protocols, and ideally, before they are put in place. In particular, we contemplate the use of these aids in the field of gastroenterology, focusing particularly on capsule endoscopy and highlighting efforts aimed at resolving the issues associated with their use when available.

Artificial intelligence and capsule endoscopy: automatic detection of enteric protruding lesions using a convolutional neural network

BACKGROUND AND AIMS

capsule endoscopy (CE) revolutionized the study of the small intestine. Nevertheless, reviewing CE images is time-consuming and prone to error. Artificial intelligence algorithms, particularly convolutional neural networks (CNN), are expected to overcome these drawbacks. Protruding lesions of the small intestine exhibit enormous morphological diversity in CE images. This study aimed to develop a CNN-based algorithm for the automatic detection small bowel protruding lesions.

METHODS

a CNN was developed using a pool of CE images containing protruding lesions or normal mucosa from 1,229 patients. A training dataset was used for the development of the model. The performance of the network was evaluated using an independent dataset, by calculating its sensitivity, specificity, accuracy, positive and negative predictive values.

RESULTS

a total of 18,625 CE images (2,830 showing protruding lesions and 15,795 normal mucosa) were included. Training and validation datasets were built with an 80 %/20 % distribution, respectively. After optimizing the architecture of the network, our model automatically detected small-bowel protruding lesions with an accuracy of 92.5 %. CNN had a sensitivity and specificity of 96.8 % and 96.5 %, respectively. The CNN analyzed the validation dataset in 53 seconds, at a rate of approximately 70 frames per second.

CONCLUSIONS

we developed an accurate CNN for the automatic detection of enteric protruding lesions with a wide range of morphologies. The development of these tools may enhance the diagnostic efficiency of CE.

Primary small bowel follicular lymphoma: the role of balloon-assisted enteroscopy in diagnosis and follow-up

An asymptomatic 38-year-old male with no significant previous medical history performed routine laboratory studies that revealed iron-deficiency anemia. Esophagogastroduodenoscopy and colonoscopy were unremarkable and he undergone videocapsule endoscopy that revealed multiple small polyps along jejunum and ileum. Double-balloon enteroscopy confirmed the presence of scattered small whitish nodules and small polyps carpeting segments of jejunal mucosal and sometimes forming conglomerates with a nodular appearance. Histopathological examination showed lamina propria expansion by neoplastic follicles, predominantly composed by small lymphoid cells that, by immunohistochemistry, showed expression of CD20, CD10 and bcl-2. Computed tomography scan of abdomen and pelvis did not reveal systemic involvement, consistent with primary small bowel follicular lymphoma. Chemotherapy was started and, at reevaluation enteroscopy, although nodular jejunal segments persisted, biopsies did not show involvement by lymphoproliferative disease, which was interpreted as complete remission. Periodic clinical and biochemical evaluation and annual enteroscopic surveillance was maintained and, after three years, local recurrence of low-grade follicular lymphoma was detected. As previously, there was no evidence of systemic involvement and the decision was to maintain close surveillance. After one year, the patient remains asymptomatic and without evidence of disease progression. This case illustrates the essential role of balloon-assisted enteroscopy for diagnosis and surveillance of primary small bowel follicular lymphoma.

Hepatic amyloidosis: a prevalence study and clinical characterization of a rare and severe disease

BACKGROUND AND AIM

Amyloidosis is a systemic disease characterized by extracellular deposition of amyloid protein, most commonly in the heart and kidney. Hepatic amyloidosis is a rare form of presentation that ranges from mild hepatomegaly and altered liver biochemical tests to acute liver failure. The aims of this study were to evaluate the prevalence of amyloidosis in patients undergoing liver biopsy and describe its main clinical characteristics and prognostic impact.

METHODS

A retrospective analysis of all patients with a histological diagnosis of hepatic amyloidosis between January 2010 and December 2019 was performed.

MAJOR RESULTS

A total of 7 patients were identified from a total of 1773 liver biopsy procedures (0.4%), with a female predominance (6/7) and median age of diagnosis of 62 years. The most common clinical manifestations included hepatomegaly (4/7), jaundice (2/7) and peripheral edema (2/7), whereas 3/7 patients were asymptomatic. Every patient presented abnormalities in liver biochemical tests, more commonly cholestasis (6/7), but also cytolysis (4/7) or hyperbilirubinemia (2/7). Abnormal imaging findings included hepatomegaly, steatosis or parenchymal heterogeneity. In most patients (5/7), other organs were involved, most commonly with nephrotic syndrome (3/7) and infiltrative cardiomyopathy (3/7). The most common type was AA amyloidosis (3/7) followed by AL amyloidosis (2/7). The 1-year mortality rate was 43% and the median survival was 24 months.

CONCLUSIONS

We report a low prevalence (0.4%) of amyloidosis among patients undergoing liver biopsy. Although rare, hepatic amyloidosis is associated with a dismal prognosis and a high index of suspicion is crucial to achieve an early diagnosis. .

Liver biopsy in inflammatory bowel disease patients with sustained abnormal liver function tests: a retrospective single-center study

Background

Inflammatory bowel disease (IBD) may be associated with a wide range of hepatobiliary manifestations. This study aimed to characterize the spectrum of hepatobiliary disorders in patients with IBD who underwent liver biopsy for sustained abnormal liver function tests (LFT).

Method

A retrospective study was performed of all patients with IBD who underwent liver biopsy between January 2010 and December 2020 for sustained abnormal LFT (at least 6-month duration).

Results

A total of 101 patients were included, mostly male (62.4%), with a mean age of 44.4±13.3 years. The most common IBD type was Crohn's disease (61.4%). Median time interval between abnormal LFT and biopsy was 14 (7-36) months. Abnormal LFT was predominantly hepatocellular in 40 patients (39.6%), cholestatic in 26 (25.7%) and mixed in 35 (34.7%). The most frequent diseases were nonalcoholic fatty liver disease (NAFLD) in 33 patients (32.7%), drug-induced liver disease (DILI) in 30 (29.7%), autoimmune hepatitis (AIH) in 13 (12.9%) and primary sclerosing cholangitis (PSC) in 13 (12.9%). Three patients had primary biliary cholangitis. Remarkably, 70 patients (69.3%) already had fibrosis by the time of liver biopsy and in 6 (5.9%) liver disease was already detected in the stage of cirrhosis.

Conclusions

Abnormal LFT in IBD patients had a wide range of etiologies and histology was often essential for reaching a correct diagnosis. NAFLD, DILI, AIH and PSC were the most common diagnoses and patients often presented in cirrhotic stage. Therefore, liver biopsy must be considered early in IBD patients with unexplained sustained abnormal LFT.

Splenic-hepatic elastography index is useful in differentiating between porto-sinusoidal vascular disease and cirrhosis in patients with portal hypertension

INTRODUCTION

In patients with portal hypertension (PH), the differential diagnosis between porto-sinusoidal vascular disease (PSVD) and cirrhosis is challenging. This study aims to evaluate the diagnostic accuracy of the SSM/LSM index in the diagnosis of PSVD.

METHODS

Prospective study of patients with PH and PSVD or cirrhosis. Transient liver and spleen elastography were performed and the ratio between spleen stiffness measurement (SSM) and liver stiffness measurement (LSM) was calculated. The relation of SSM/LSM with the diagnosis of PSVD was evaluated.

RESULTS

Forty-four patients with PSVD and 44 patients with cirrhosis were evaluated. Median age was 57.5 (IQR 49.0-64.5) years, 66.3% were males. In patients with PSVD, median SSM was 59.4 (33.5-77.7) kPa, median LSM was 6.2 (5.2-10.2) kPa and median SSM/LSM was 5.62 (3.15-9.68). In patients with cirrhosis, median SSM was 47.3 (24.3-60.3) kPa, median LSM was 27.8 (17.7-53.9) kPa and median SSM/LSM was 1.55 (1.06-3.24). The SSM/LSM AUROC was 0.940 (p<0.001). Using 2 as a cut-off, we obtained good sensitivity (86.5%), specificity (92.7%), and accuracy (89.7%) for the diagnosis of PSVD.

CONCLUSION

The SSM/LSM index is useful in the differential diagnosis between liver cirrhosis and PSVD. Using the cut-off of 2 we achieved a good sensitivity and specificity for diagnosing PSVD.

Peri-operative management of a patient with an ectodermal dysplasia (Rapp-Hodgkin) syndrome

We present the case of a 41-year-old man with Rapp-Hodgkin syndrome who underwent nasal septum deviation surgery under general anaesthesia. This syndrome is rare, with approximately 70 cases reported worldwide. It is one of a group of ectodermal dysplasia syndromes and results from the aberrant development of ectoderm during fetal development. Some of the clinical features may affect anaesthetic management. The most important considerations are potentially difficult airway management, the need for meticulous temperature control, and the importance of skin protection. This case was uneventful, but as there are few case reports on the management of patients with ectodermal dysplasia syndromes undergoing anaesthesia this report contributes useful knowledge. The pathogenesis and clinical features of Rapp-Hodgkin syndrome and the anaesthetic management for this patient are described.

Artificial intelligence and colon capsule endoscopy: Automatic detection of ulcers and erosions using a convolutional neural network

BACKGROUND AND AIM

Colon capsule endoscopy (CCE) has become a minimally invasive alternative for conventional colonoscopy. Nevertheless, each CCE exam produces between 50 000 and 100 000 frames, making its analysis time-consuming and prone to errors. Convolutional neural networks (CNNs) are a type of artificial intelligence (AI) architecture with high performance in image analysis. This study aims to develop a CNN model for the identification of colonic ulcers and erosions in CCE images.

METHODS

A CNN model was designed using a database of CCE images. A total of 124 CCE exams performed between 2010 and 2020 in two centers were reviewed. For CNN development, a total of 37 319 images were extracted, 33 749 showing normal colonic mucosa and 3570 showing colonic ulcers and erosions. Datasets for CNN training, validation, and testing were created. The performance of the algorithm was evaluated regarding its sensitivity, specificity, positive and negative predictive values, accuracy, and area under the curve.

RESULTS

The network had a sensitivity of 96.9% and a specificity of 99.9% specific for the detection of colonic ulcers and erosions. The algorithm had an overall accuracy of 99.6%. The area under the curve was 1.00. The CNN had an image processing capacity of 90 frames per second.

CONCLUSIONS

The developed algorithm is the first CNN-based model to accurately detect ulcers and erosions in CCE images, also providing a good image processing performance. The development of these AI systems may contribute to improve both the diagnostic and time efficiency of CCE exams, facilitating CCE adoption to routine clinical practice.

Bronchial-biliary fistula secondary to cholangiocarcinoma: long-term efficacy of biliary self-expandable metal stent

A 77-year-old female with previous medical history of non-cirrhotic chronic hepatitis B and hepatocellular carcinoma treated with sequential partial hepatectomy followed by transarterial chemoembolization complained of pruritus and jaundice. Magnetic resonance cholangiopancreatography revealed a peri-hilar ill-defined stenosing lesion suggestive of malignancy. Endoscopic retrograde cholangiopancreatography with cholangioscopy confirmed a circumferential peri-hilar stenosis with fragile mucosa and tortuous dilated vessels and biopsies of this area were consistent cholangiocarcinoma. After 3 months, she presented with new-onset dyspnea and bilioptysis and abdominal computed tomography revealed a bronchial-biliary fistula. ERCP was performed to place a self-expandable metal stent in the biliary tract, which resulted in rapid clinical improvement. The patient has been followed for 2 years and remains globally stable with two episodes of worsening of bilioptysis secondary to stent obstruction by lithiasis that were easily resolved with Fogarty balloon-assisted extraction, with rapid improvement. This case demonstrates the long-term efficacy of endoscopic biliary drainage with self-expandable metallic stent for bronchial-biliary fistula in the setting of cholangiocarcinoma. .

Deep Learning for Automatic Differentiation of Mucinous versus Non-Mucinous Pancreatic Cystic Lesions: A Pilot Study

Endoscopic ultrasound (EUS) morphology can aid in the discrimination between mucinous and non-mucinous pancreatic cystic lesions (PCLs) but has several limitations that can be overcome by artificial intelligence. We developed a convolutional neural network (CNN) algorithm for the automatic diagnosis of mucinous PCLs. Images retrieved from videos of EUS examinations for PCL characterization were used for the development, training, and validation of a CNN for mucinous cyst diagnosis. The performance of the CNN was measured calculating the area under the receiving operator characteristic curve (AUC), sensitivity, specificity, and positive and negative predictive values. A total of 5505 images from 28 pancreatic cysts were used (3725 from mucinous lesions and 1780 from non-mucinous cysts). The model had an overall accuracy of 98.5%, sensitivity of 98.3%, specificity of 98.9% and AUC of 1. The image processing speed of the CNN was 7.2 ms per frame. We developed a deep learning algorithm that differentiated mucinous and non-mucinous cysts with high accuracy. The present CNN may constitute an important tool to help risk stratify PCLs.

Artificial Intelligence and Device-Assisted Enteroscopy: Automatic Detection of Enteric Protruding Lesions Using a Convolutional Neural Network

INTRODUCTION

Device-assisted enteroscopy (DAE) plays a major role in the investigation and endoscopic treatment of small bowel diseases. Recently, the implementation of artificial intelligence (AI) algorithms to gastroenterology has been the focus of great interest. Our aim was to develop an AI model for the automatic detection of protruding lesions in DAE images.

METHODS

A deep learning algorithm based on a convolutional neural network was designed. Each frame was evaluated for the presence of enteric protruding lesions. The area under the curve, sensitivity, specificity, and positive and negative predictive values were used to assess the performance of the convolutional neural network.

RESULTS

A total of 7,925 images from 72 patients were included. Our model had a sensitivity and specificity of 97.0% and 97.4%, respectively. The area under the curve was 1.00.

DISCUSSION

Our model was able to efficiently detect enteric protruding lesions. The development of AI tools may enhance the diagnostic capacity of deep enteroscopy techniques.