Abstract
Background
Artificial intelligence (AI) systems, in particular neural networks (NN), have been developed to detect lesions in wireless capsule endoscopy (WCE) in patients with Crohn’s ...disease (CD) in order to ease their reading. Despite acceptable performance on individual images, these tools do not analyze entire videos and require a large amount of labeled data. Active learning (AL) methods have already been used, proving their ability to limit labeling cost without reducing NN performance. The aim of this study was to evaluate several AL models in order to train an AI system able to differentiate normal and pathological entire videos with a limited labeled dataset.
Methods
Firstly, 76 WCEs of CD patients were used to automate the detection of the first and last image of the small bowel. They were identified by an expert and then used for training, validation and test of a NN. The model associated with the Viterbi algorithm predicted boundaries location and was compared with expert annotations.
Secondly, 4286 images from the Crohn-IPI database¹ were used to train another NN to classify pathological or normal images. An image was considered pathological if it contained erythema, oedema, ulceration or stenosis. Four AL models with several selection criteria² were tested on a limited labeled dataset and their classification performance (normal/pathological image) was evaluated.
Finally, 42 WCEs were used to assess performance of a pre-trained AL model to predict the normal or pathological nature of entire videos.
Results
Overall, the model detected the pylorus compared with experts with a median deviation (IQR) of 3 images (66.5) and the last ileal image with a median deviation of 98 images (336). A history of ileocolic resection non-significantly impaired the last ileal image location, with a median difference of 203 images (812, p=0.59).
None of the four AL models showed superiority in classifying pathological or normal image, with a sensitivity of 60% and a specificity of 98%.
Finally, the AL pre-trained NN assessed the pathological/normal status of entire videos with a sensitivity of 83% and a specificity of 50%.
Conclusion
Our study suggests the feasibility and efficiency of NN to identify the first and last image of the small bowel on CD patients WCE. Performances of the four AL models were similar but remained sufficient given the small amount of labeled data used. Prediction of the pathological/normal status of entire videos by AI systems seems feasible, but requires optimization to obtain acceptable results for current practice.
1. De Maissin A et al. Multi-expert annotation of Crohn's disease(...) neural network, Endosc IntOpen(2021)
2. Wang K et al. Cost-Effective Active Learning for Deep Image Classification, IEEE (...) Technology(2017)
Objective control of intestinal inflammation during inflammatory bowel disease (IBD) is becoming the main driver for medical treatment. However, the monitoring tools-related burden remains poorly ...investigated. We aimed to evaluate their comparative acceptability and utility according to patients with IBD.
After a preliminary phase, the final questionnaire encompassing self-administered and physician questionnaires was prospectively and consecutively submitted to 916 patients with IBD from 20 public and private centers. Acceptability and utility visual analog scales (VAS) were expressed as median with interquartile range.
Regarding the group of patients with Crohn's disease (n = 618), venipuncture (VAS = 9.3 8.8-9.7) and ultrasonography (VAS = 9.3 8.7-9.7) were the most acceptable tools (P < 0.0001, for each comparison), whereas rectosigmoidoscopy was the least acceptable tool (VAS = 4.4 1.2-7.3) (P < 0.0001, for each comparison). Wireless capsule endoscopy (VAS = 8.5 5.2-9.3), magnetic resonance enterocolonography (VAS = 8.0 5.0-9.2), and stools collection (VAS = 7.7 4.6-9.3) were more acceptable than colonoscopy (VAS = 6.7 4.3-8.9) (P < 0.0001, for each comparison). The acceptability was assessed in 298 patients with ulcerative colitis for venipuncture (VAS = 9.4 8.8-9.7), stools collection (VAS = 8.1 5.7-9.4), colonoscopy (VAS = 7.5 4.7-9.2), and rectosigmoidoscopy (VAS = 6.7 2.8-9.1); (P < 0.001 for each comparison). All monitoring tools were considered as highly useful by patients with IBD. Decreased acceptability was related to embarrassment for the collection/transport of stools (60.7%), bowel cleansing (76.3%) for colonoscopy, abdominal discomfort (51.3%) and rectal enema (36.6%) for rectosigmoidoscopy, bowel distension (48.3%) for magnetic resonance enterocolonography, and potential capsule retention (21.4%) for wireless capsule endoscopy.
Among the IBD monitoring tools, endoscopy demonstrated the lowest acceptability supporting the development of alternative modalities. Patients' information and examination conditions should be improved to ensure proper monitoring adherence.
Background & Aims: The aim of this study was to evaluate the usefulness of short-term infliximab combined with azathioprine (AZA) or 6-mercaptopurine (6-MP) in steroid-dependent Crohn’s disease ...patients. Methods: Patients with active disease despite prednisone given for more than 6 months were eligible and were stratified as follows: the failure stratum consisted of patients receiving AZA/6-MP at a stable dose for more than 6 months, and the naive stratum consisted of patients not treated previously with AZA/6-MP. Patients were randomized to infliximab 5 mg/kg or placebo at weeks 0, 2, and 6. All patients were treated with AZA/6-MP maintained at a stable dose throughout the 52 weeks of the study. The primary end point was remission off steroids at week 24. Results: Among the 113 enrolled patients (55 in the failure stratum), 57 were assigned to infliximab. At week 24, the success rate (intent-to-treat analysis) was higher in the infliximab group than in the placebo group (57% vs 29%; P = .003); at weeks 12 and 52, the corresponding rates were 75% vs 38% (P < .001) and 40% vs 22% (P = .04), respectively. In each stratum, the success rate was significantly higher in the infliximab group at weeks 12 and 24, and a trend was found at week 52. In the failure stratum, only 27% of the patients in the infliximab group were still in remission off steroids, compared with 52% in the naive stratum. Steroid resistance was less common and the cumulative dose of prednisone was lower in the infliximab group. Conclusions: Infliximab plus AZA/6-MP is more effective than AZA/6-MP alone in steroid-dependent Crohn’s disease patients.
La compréhension des mécanismes physiopathologiques des maladies inflammatoires chroniques intestinales (MICI) a permis le développement récent de nouvelles thérapies, dites biologiques.
...L'infliximab, chef de file de ces thérapies est largement utilisé en cas de résistance aux thérapeutiques de première intention ou en cas de corticodépendance. De nouveaux anti-TNF-α comme l'adalimumab ou le certolizumab renforcent l'arsenal thérapeutique et permettront la poursuite des thérapies ciblées contre le TNF-α malgré les résistances ou les intolérances à l'infliximab. Le développement de certaines thérapies a été abandonné par manque d'efficacité ou retardé par l'apparition d'effets indésirables comme les cas de leucoencéphalite multifocale progressive décrits sous natalizumab. Enfin, certains traitements comme le GM-CSF visant à renforcer les défenses immunitaires des patients atteints de maladie de Crohn apparaissent prometteurs.
La remarquable efficacité de ces nouvelles molécules permettra de modifier la prise en charge des patients atteints de maladie inflammatoire chronique intestinale et de diminuer le recours aux corticoïdes à forte dose dont la tolérance et l'acceptabilité par les malades sont mauvaises.
Advances in the understanding of inflammatory bowel disease (IBD) pathophysiological mechanisms in the last few years have allowed the development of novel therapies such as biologic therapies. Theoretically, biologic therapies represent a more specific management of IBD with fewer effects.
Currently, infliximab is the only effective and widely accepted biologic therapy for the treatment of Crohn disease after the conventional therapies. Others anti-TNF therapies such as adalimumab or certolizumab will be soon an alternative treatment notably for patients with allergic reactions to infliximab and for those with lost of response because of anti-infliximab antibody development. Anti-integrin α4 therapies have been delayed by three progressive multifocal leukoencephalopathy cases. Immunostimulating therapy may be highly relevant in the future with granulocyte–monocyte colony-stimulating factor.
Efficacy of these new therapies will modify therapeutics of Crohn's disease and ulcerative colitis and in particular decrease the use of corticosteroids, which are not well tolerated by the patients.
Few studies have been conducted addressing the safety of thiopurine treatment in pregnant women with inflammatory bowel disease (IBD). The aim of this study was to evaluate the pregnancy outcome of ...women with IBD who have been exposed to thiopurines.
215 pregnancies in 204 women were registered and documented in the CESAME cohort between May 2004 and October 2007. Physicians documented the following information from the women: last menstrual date, delivery term, details of pregnancy outcome, prematurity, birth weight and height, congenital abnormalities, medication history during each trimester, smoking history and alcohol ingestion. Data were compared between three groups: women exposed to thiopurines (group A), women receiving a drug other than thiopurines (group B) and women not receiving any medication (group C).
Mean age at pregnancy was 28.3 years. 75.7% of the women had Crohn's disease and 21.8% had ulcerative colitis, with a mean disease duration of 6.8 years at inclusion. Of the 215 pregnancies, there were 138 births (142 newborns), and the mean birth weight was 3135 g. There were 86 pregnancies in group A, 84 in group B and 45 in group C. Interrupted pregnancies occurred in 36% of patients enrolled in group A, 33% of patients enrolled in group B, and 40% of patients enrolled in group C; congenital abnormalities arose in 3.6% of group A cases and 7.1% of group B cases. No significant differences were found between the three groups in overall pregnancy outcome.
The results obtained from this cohort indicate that thiopurine use during pregnancy is not associated with increased risks, including congenital abnormalities.
Abstract
Background
Deep remission has become the target in the management of Crohn’s disease (CD). The role of histology in the management of CD has not yet been established. The aim of this work ...was to study the impact of histological inflammation on the risk of clinical relapse in patients with CD in clinical and endoscopic remission and to study the correlation between histology and endoscopic scores and biomarkers.
Methods
All patients included in STORI (1) were eligible for inclusion. Fecal calproptectin and colonoscopy with CDEIS calculation were performed at inclusion. Two biopsies, taken either from the most inflammed macroscopic area or from an area with previous inflammation in case of mucosal healing (MH) were prospectively performed. MH was defined by the absence of ulcers. Different histological scores were calculated by 2 independent pathologists: Robart, Geboes, modified Geboes, Nancy, and DCA-IBD scores. Histological remission was defined by Nancy=0 ; Geboes ≤2 ; modified Geboes of grade 0 ; Robarts ≤3, IBD-DCA of C≤1 and A=0. Clinical relapse was prospectively evaluated in the STORI trial as CDAI>250 or CDAI between 150 and 250 with an increase of 70 points during 2 successive weeks compared to the inclusion visit.
(1) Louis E et al .Gastroenterology 2012
Results
Eighty biopsies (22 ileal and 58 colonic) were available from 76/115 patients included in STORI. Among the 45 patients with MH, 30 had colonic biopsies, 11 had ileal biopsies and 4 had biopsies in both locations. Among colonic biopsies 68% had histological remission according to Nancy and modified Geboes scores, 65% according to DCA-IBD and Geboes scores and 71% according to Robarts score. Among ileal biopsies 53% had histological remission according to Nancy score, 47% according to Modified Geboes, Geboes and DCA-IBD scores, and 67% according to the Robarts score. Thirty-five patients (46%) experienced a clinical relapse during the follow-up including 16/45 (36%) and 19/31 (61%) in the group with MH and without MH respectively. Histological characteristics and histological scores were not predictive of clinical relapse. In the total cohort (n=76), the CDEIS score was correlated to Nancy (p<0,001, r=0,43), Geboes (p<0,001, r=0,47) and Robarts (p<0,001, r=0,45) scores. Fecal calprotectin was correlated to Nancy (p<0,001, r=0,48), Geboes (p<0,001, r=0,45) and Robarts (p<0,001, r=0,46) scores. The histological scores of Nancy, Robarts and Geboes were highly correlated (p<0,0001 ; r > 0,9).
Conclusion
Although correlated with endoscopy and biomarkers, histology was not predictive of clinical relapse in CD patients in clinical and endoscopic remission. These findings do not support the use of histology to predict clinical relapse in CD in clinical practice.
Abstract
Background
The intestinal epithelium can be easily disrupted during gut inflammation as seen in inflammatory bowel diseases (IBD) such as ulcerative colitis (UC) or Crohn’s disease (CD). ...Aetiology of IBD is still not fully understood, but recent evidences suggest that the intestinal epithelium might play a major role in the development and perpetuation of IBD. In fact, disturbances in mechanisms that control the homeostasis, protection and repair of intestinal epithelial cells can lead to increased intestinal permeability causing deregulated immune response to the commensal gut microbiota and ultimately chronic intestinal inflammation. The cytokines IL-22 and IL-17 are highly produced in the inflammatory mucosa of IBD patients. In rodent models of colitis, these two cytokines showed synergistic and protective roles during gut inflammation, by reinforcing epithelial barrier function. We have shown that IL-22 binding protein (IL-22BP), the soluble and specific inhibitor of IL-22, is also increased during IBD and could therefore hamper the protective actions of IL-22.
Methods
To further explore this hypothesis, we set up ex vivo cultures of colonic biopsies from patients with active CD and UC and analysed the expression and regulation of IL-22-dependent genes that may be controlled by IL-22BP.
Results
We first observed that, as previously described by others, the antimicrobial peptides (AMPs) BD2, BD3 and LNC2, known targets of IL-22, were induced at a lower level in the inflammatory mucosa of CD than of UC patients. We then demonstrated that this defect in AMPs expression in CD was reversed by ex vivo stimulation with IL-22 and IL-17, and identified IL-22 vs. IL-17-dependent as well as IL-22+IL-17 synergistic responses. Furthermore, we showed that the addition of IL-22BP to the culture medium blocked the induction of IL-22-dependent genes.
Conclusions
Our data strongly suggest that the defective AMPs production observed in CD might be related to lack of IL-22 and IL-17 actions on epithelial cells. We propose that the selective and transient blockade of IL-22BP could represent an interesting therapeutic strategy to unleash the protective effects of locally-produced IL-22 during flares in CD.
Enteric glial cells (EGCs) are important regulators of intestinal epithelial barrier (IEB) functions. EGC-derived S-nitrosoglutathione (GSNO) has been shown to regulate IEB permeability. Whether EGCs ...and GSNO protect the IEB during infectious insult by pathogens such as Shigella flexneri is not known.
S flexneri effects were characterised using in vitro coculture models of Caco-2 cells and EGCs (or GSNO), ex vivo human colonic mucosa, and in vivo ligated rabbit intestinal loops. The effect of EGCs on S flexneri-induced changes in the invasion area and the inflammatory response were analysed by combining immunohistochemical, ELISA and PCR methods. Expression of small G-proteins was analysed by western blot. Expression of ZO-1 and localisation of bacteria were analysed by fluorescence microscopy.
EGCs significantly reduced barrier lesions and inflammatory response induced by S flexneri in Caco-2 monolayers. The EGC-mediated effects were reproduced by GSNO, but not by reduced glutathione, and pharmacological inhibition of pathways involved in GSNO synthesis reduced EGC protecting effects. Furthermore, expression of Cdc42 and phospho-PAK in Caco-2 monolayers was significantly reduced in the presence of EGCs or GSNO. In addition, changes in ZO-1 expression and distribution induced by S flexneri were prevented by EGCs and GSNO. Finally, GSNO reduced S flexneri-induced lesions of the IEB in human mucosal colonic explants and in a rabbit model of shigellosis.
These results highlight a major protective function of EGCs and GSNO in the IEB against S flexneri attack. Consequently, this study lays the scientific basis for using GSNO to reduce barrier susceptibility to infectious or inflammatory challenge.
Abstract
Background
There is no biomarker to guide first-line of biotherapy in ulcerative colitis (UC). The main objective of this multicenter pilot study was to demonstrate the feasibility of ...identifying a4b7 and/or TNF-a expressing cells by dual-band confocal endomicroscopy (CLE) into the mucosa to predict the response to vedolizumab as first-line biotherapy. NCT02878083.
Methods
Patients with moderate to severe UC, naïve of biotherapy were prospectively included and received vedolizumab. Patients were evaluated clinically, biologically, endoscopically and histologically at week 22. Nonresponding patients were subsequently treated by adalimumab and were evaluated at week 30. Prior to the initiation of vedolizumab, during endoscopy, 8 biopsies were taken for extemporaneous and ex vivo analysis by CLE, histology and confocal microscopy. Fresh biopsies were treated with vedolizumab coupled with FITC and adalimumab coupled with Alexa-Fluor to analyze their respective fluorescence (figure) in one step with the Dual-band CLE system (Mauna Kea Technologies, France). The number and the surface of positive stained areas visualized by CLE were counted and compared between responders and non-responders. Blood samples were collected before each infusion and were used to analyze the immune cells (LT, LB, NK, monocytes) by FACS and vedolizumab trough levels. Frozen biopsies were analyzed by confocal microscopy (Opal Akoya multiplexing; Tyramide signal amplification) after labeling by an anti-a4b7 antibody.
Results
19 patients out of the 20 initially planned were included. Clinical remission, endoscopic improvement (Mayo ≤ 1), endoscopic remission (Mayo 0) at W22 were respectively 58%, 58% and 45%. 18 out of 19 patients (95%) could be analyzed by CLE. The number and surface of vedolizumab-FITC stained areas were numerically superior in responder patients at week22; by contrast, no difference was found for the areas marked by adalimumab-Alexa-fluor. A threshold value of 6 areas marked in CLE by the vedolizumab-FITC at week 0 discriminated responders and non-responders at week 22 with a Se of 78% and a Sp of 85%. No difference of circulating cell population has been identified between responders and non-responders and the vedolizumab trough levels were not predictive of the response. In confocal microscopy, 15 patients have been analyzed; the percentage of a4b7-binding cells was not different between responders (n=9) and non-responders (n=6)
Conclusion
This study confirms the feasibility of extemporaneous identification by dual-band CLE of the mucosal cells expressing a4b7 and/or TNF-a. Results suggest that only the proportion of cells binding vedolizumab-FITC in the mucosa was indicative of the response to vedolizumab