Exabis Library

Educational objectives:
1. To give an overview of the main cutaneous manifestation of IBD 
2. To understand the importance of the clinico-pathological correlation for their diagnosis 

Crohn’s disease (CD) is a lifelong chronic illness with recurrent relapsing and remitting disease course that requires close follow ups and reassessments of disease status as well as screening for complications throughout a patient’s lifetime. Imaging plays a crucial role in the diagnosis and evaluation of CD. Currently, different imaging modalities can be used over the disease course of a patient’s lifetime, from the diagnosis of the disease, to determining the extent of intestinal involvement, monitoring for disease activity, and evaluating for CD related complications. Intestinal Ultrasound (IUS) is a non-invasive, radiation-free, and safe useful imaging tool that can be used in the diagnosis and management of crohn’s disease. 

Educational Objectives:

1. To emphasize on the increased rule of imaging in IBD management.
2. To review the rule of Intestinal ultrasound (IUS) in detecting Crohn’s disease complications:
   1. Strictures.
   2. Abdominal fistulas.
   3. Inflammatory Masses.
3. To review the latest guidelines statements for the use of intestinal Ultrasound in detecting crohn’s disease complications.
4. Quick overview on the innovated tools to improve the accuracy of intestinal ultrasound in detecting crohn’s disease complications.

Interactive cases will be presented to illustrate IUS diagnosis of UC complications, such as acute severe colitis, including monitoring of response to treatment, perforation, chronic structural damage in UC, infectious colitis, pseudopolips and lymphoma.

Background

Capsule endoscopy (CE) is the gold-standard for the evaluation of the enteric mucosa in patients with suspected or known inflammatory bowel disease, particularly Crohn’s disease. Ulcers and erosions of the small bowel are common findings and their identification in CE is paramount for an accurate disease stratification.

Several artificial intelligence (AI) algorithms have been developed to aid endoscopists to detect lesions in different endoscopic modalities. With this project we intend to develop and test an AI algorithm for the automatic identification of ulcers and erosions in the small bowel mucosa.

Methods

A total of 2565 CE exams from two different centers (1483 from São João University Hospital and 1082 from ManopH Gastroenterology Clinic) were used to develop the Convolutional Neural Network (CNN). 55320 frames of the enteric mucosa were obtained, 18396 containing enteric ulcers and erosions, and 36924 containing normal mucosa. 90% of the frames were used to develop the training dataset and 10% were used to test the network. The patients included on the training dataset were excluded from the testing dataset. This patient split brings the technology performance closer to that of a real-life setting. The output provided by the CNN was compared to the classification provided by a consensus of experts.

Results

Our model was able to automatically detect ulcers and erosions in the enteric mucosa with an accuracy of 93.2%, sensitivity of 90.4% and a specificity of 93.9%. The mean processing time for the validation dataset was 29 seconds (approximately 306 frames/second).
An example of the output obtained after the network application can be seen in Figure 1.

Conclusion

The authors developed a CNN for the automatic identification of enteric ulcers and erosions in CE videos and tested it in AI naïve patients. This represents an evolution in the technology readiness level into a real-life clinical setting, that will surely improve the diagnostic yield of CE exams, which will ultimately translate into better patient care.

Early prediction of intravenous corticosteroid (IVCS) resistance in Acute severe ulcerative colitis (ASUC) patients could reduce costs and delay in rescue therapy. However, most prediction models for ASUC were at high risk of bias with a lack of external validation. This study aims to construct and validate a model that accurately predicts IVCS resistance using various statistical methods.

A retrospective cohort of patients who were diagnosed with ASUC and had undergone IVCS treatment between March 2012 to January 2020 was established. Predictors evaluated included age, gender, race, medications before admission, infections, and laboratory data at baseline and during IVCS treatment, and endoscopic outcomes relied on blinded centralized endoscopy reading. The LASSO regression was used in feature selection for multivariate logistic regression model. Models based on machine learning methods (decision tree and random forest [RF]) were also constructed. Internal validity was confirmed and model performances were compared. External validation was conducted using data using an independent cohort from a tertiary referral centre.

A total of 129 patients were included in the derivation cohort. During index hospitalization, 102 (79.1%) responded to IVCS, and 27 (20.9%) failed; 16 patients underwent colectomy, 6 received cyclosporin, and 5 succeeded with IFX as rescue therapy. Ulcerative Colitis Endoscopic Index of Severity (UCEIS; odds ratio [OR] 5.39, 95% confidence interval [CI] 2.52-14.0, p<0.001) and C-reactive protein (CRP) level on the third day (OR 1.05, 95% CI 1.03-1.08, p<0.001) were selected by LASSO regression and identified as the only two independent predictors of IVCS resistance in logistic regression. The decision tree model identified a UCEIS higher than 6.5 points and CRP level at day 3 higher than 33.57 mg/dL as the proxy for IVCS resistance. UCEIS and CRP level at day 3 were also the most important predictors in the RF model. Areas under the curve receiver operating characteristic (AUC) of logistic model, decision tree model, and RF model were 0.64 (95% CI 0.49-0.80), 0.81 (95% CI 0.71-0.90), and 0.88 (95% CI 0.82-0.95), respectively. A validation cohort of 65 ASUC patients were established, and the AUC of the models in external validation were 0.57 (95% CI 0.45-0.70), 0.70 (95% CI 0.61-0.80), and 0.71 (95% CI 0.48-0.94), respectively.

In patients with ASUC, UCEIS and CRP level at day 3 of IVCS treatment appeared to allow the prompt prediction of likely IVCS nonresponders. Machine learning-based models outperformed the traditional statistical model in the prediction. The models may aid therapeutic decision-making in ASUC patients.

In developing a patient-reported outcome (PRO) for pediatric ulcerative colitis (UC) with guidance from FDA and EMA, 8 items were previously selected based on 79 concept elicitation interviews. An observer RO (ObsRO) was determined to be required for children younger than 8 years. Here, we aimed to finalize the included items and to validate the TUMMY-UC for its psychometric properties.

The structure and exact wording of the PRO and the ObsRO versions were determined by cognitive debriefing interviews with children and their caregivers. Weights were assigned to each item based on ranking of importance. Then, in a prospective multicenter study, children with UC between 2-18 years who either underwent colonoscopy or provided stool for calprotectin completed the TUMMY-UC during 4 consecutive days, as well as 7 and 21 days thereafter for evaluating reliability and responsiveness. Construct and discriminative validity were assessed by different measures of disease severity and quality of life (QOL).  

In an iterative process of 129 cognitive interviews, the exact wording of the TUMMY-UC was determined. The PRO and ObsRO were formatted with identical structure to ensure conceptual equivalence for incorporating into one score. 71 children were included in the validation study (39 with colonoscopy and 32 with calprotectin; age 12.3±4.1 years, 26 (36%) in remission, 20 (29%) with moderate-severe disease). There was excellent reliability in the repeated day assessments (ICC 0.93 (0.88-0.96); p<0.001) and after 1 week in those judged as unchanged (0.90 (0.81-0.95); p<0.001). The TUMMY-UC total score had moderate to strong correlations with all constructs of disease severity: r=0.64 with UC Endoscopic Index of Severity (UCEIS, Figure 1), r=0.66 with IMPACT QOL questionnaire, r=0.43 with calprotectin, r=0.82 with the PUCAI, r=0.76 with patient/caregiver global assessment, r=0.5 with CRP, and r=-0.36 with albumin (all p<0.015). There was a slight superiority to combining TUMMY-UC scores of two consecutive days. The index had excellent discrimination of disease activity categories (figure 2) with a score<9 defining remission (Sen=93%, Spec=84%, AUROC=0.95 (95%CI 0.89-0.99). Showing high responsiveness, the DTUMMY-UC differentiated well between children who improved, worsened or remained unchanged after 3 weeks (Figure 3).The best cutoff of the TUMMY-UC to define response was a change of ≥10 points (AUROC 0.93 (95%CI 0.86-0.99)).

The TUMMY-UC, constructed from a PRO and ObsRO versions for children 8-18 and 2-7 years, respectively, is a reliable, valid and responsive index which can be now used in clinical practice and as an outcome measure in clinical trials.

Humans are colonized by complex microbial communities which contribute to physiological processes in the host. The communication between microbes and host is crucial to maintain the homeostasis and gut health. Disruption in the microbiome composition leads to increased inflammation and appearance of diseases, such as inflammatory bowel disease (IBD). Interspecies interaction prediction combined with gene expression patterns on individual cell level by single-cell omics data reveals a new insight into the molecular background of cell-type specific host-microbe interactions.

Previously we developed the MicrobioLink pipeline (Andrighetti et al, Cells, 2020), an in silico microbe-host protein-protein interaction prediction algorithm. Here, we implemented a computational workflow based on MicrobioLink to predict and compare the cell-specific effects of a commensal bacteria in healthy and diseased conditions using a publicly available single-cell RNAseq dataset (Smilie et al, Cell, 2019) from colon biopsies describing 51 cell types  - including fibroblasts, epithelial and immune cells - in healthy, non-inflamed and inflamed ulcerative colitis (UC). With functional analysis, microbe-affected processes have been discovered, while reliable network biology resources, such as OmniPath and Reactome, were used to identify the direct mechanism of action of the bacterial molecules.

Demonstrating the applicability of the new computational workflow, we analysed the effect of a common gut commensal bacteria - Bacteroides thetaiotaomicron (Bt) - on human immune cells focusing on the Toll-like receptor (TLR) signalling. We found that extracellular vesicles (EVs) secreted by Bt may be able to modulate the TLR pathway intracellularly. The analysis highlighted that Bt targets differ among cells and between the same cells in healthy versus UC conditions. The in silico findings were validated in EV-monocyte co-cultures demonstrating the requirement for TLR4 and Toll-interleukin-1 receptor domain-containing adaptor protein (TIRAP) in EV-elicited NF-kB activation.

The current pipeline offers potentially interesting connection points and detailed mechanistic insight containing mechanistic information about microbe-host interactions. This information can be tested and harnessed to understand better how microbial proteins may be of therapeutic value in inflammatory diseases, such as IBD.

Previous studies have described machine learning (ML) models to predict how human readers would score disease activity in UC using the endoscopic Mayo Score (eMS). So far, none employed deep human annotation that considers all the endoscopic features making up the eMS.  Here we report the results of an ML model that is trained on eMS features using centrally read endoscopies.

793 full-length videos were obtained from 249 patients with UC who participated in NCT02589665, a phase 2 trial with mirikizumab in patients with UC and associated with centrally read (single reader) eMS (CReMS) as the primary dataset. After cleaning for usable frames, the data were split into training, validation and testing subsets. The ML workflow consisted of annotation, segmentation, and classification (e.g., erosions, ulcers, erythema, vascular pattern, and bleeding). Human image classification and segmentation with bounding boxes and was subjected to quality control adjudicated by one of three IBD specialists, generating more than 60,000 eMS-relevant annotation labels. The model was evaluated on a test set of 147 videos using the CReMS, and a consensus set of 94 test videos, where CReMS and annotator reported eMS (AReMS) were in agreement without adjudication. The primary objective of the model was a categorical prediction of endoscopically inactive disease (eMS 0 & 1) compared with active disease (eMS 2 & 3). The secondary objectives of the model were to predict endoscopic healing (eMS 0) and to predict severe disease (eMS 3).

The model performances are in Table 1. On the full test set of 147 videos, the model predicted inactive disease compared with active disease with an accuracy of 84%, positive predictive value (PPV) of 80%, and negative predictive value (NPV) of 85%. In the subset of 94 videos with CreMS and AReMS consensus, the model predicted inactive disease compared with active disease with an accuracy of 89%, PPV 87%, and NPV of 90%. In this same subset, the model predicted endoscopic healing and severe disease with an accuracy of 95% and 85%, PPVs of 86% and 82% and NPVs of 95% and 87%, respectively. For the secondary objectives in the full set of 147 videos, the model predicted endoscopic healing and severe disease with an accuracy of 90% and 80%, PPVs of 44% and 86%, and NPVs of 95% and 86%, respectively.

We have developed a ML predictive model of the eMS in UC using centrally read videos and demonstrate excellent distinction between active and inactive disease, and clear discrimination between other levels of endoscopic activity. We propose that this unique ML approach to endoscopic assessment be considered as a substitute to human central reading in future clinical trials.

1. To understand normal physiology of the gastrointestinal tract
2. To review the etiopathogenesis of Inflammatory Bowel Diseases
3. To emphasize the different modalities to diagnose Inflammatory Bowel Diseases
4. To have an overview of the most commonly used clinical and endoscopic activity scores for Inflammatory Bowel Diseases

Although extra-intestinal manifestations are common, inflammatory bowel diseases (IBD) typically affect the intestine. Where ulcerative colitis (UC) is limited to the colon, Crohn’s disease (CD) may involve all parts of the gastrointestinal tract (mouth, oesophagus, stomach, small intestine, colon and rectum). Consequently, the function of all these segments may be compromised.

Although the exact etiopathogenesis of IBD has not been unravelled, the prevailing model states that IBD is driven by environmental factors in genetically susceptible individuals, resulting in a dysregulated immune response towards the intestinal microbiome.

Besides a good clinical history and physical examination, several diagnostic tools will help the physician to diagnose IBD. These tools include lab test (both blood and faeces), radiological examination (ultrasound, CT and MR scan), and endoscopy with biopsies for histological examination.

Clinical disease activity of CD and UC, are most commonly assessed using the Harvey-Bradshaw index and the Mayo score, respectively. However, patient reported outcomes become more frequently implemented. Also endoscopic disease activity indices have been introduced in daily clinical care. For CD, the Crohn’s disease endoscopic index of severity (CDEIS), the simple endoscopic score for Crohn’s disease (SES-CD), and the Rutgeerts score are used. For UC, the Mayo score and the ulcerative colitis endoscopic index of severity (UCEIS) are employed. C-reactive protein (CRP) and faecal calprotectin could be regarded as surrogate markers for endoscopic disease activity, but their accuracy is not optimal.