Development and validation of a deep learning-based approach to predict the Mayo endoscopic score of ulcerative colitis
Ontology highlight
ABSTRACT: Background:
The ulcerative colitis (UC) Mayo endoscopy score is a useful tool for
SUBMITTER:
PROVIDER: S-EPMC10214058 | biostudies-literature | 2023 Jan
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Background and aimsEndoscopic assessment is a co-primary end point in inflammatory bowel disease registration trials, yet it is subject to inter- and intraobserver variability. We present an original machine learning approach to Endoscopic Mayo Score (eMS) prediction in ulcerative colitis and report the model's performance in differentiating key levels of endoscopic disease activity on full-length procedure videos.MethodsSeven hundred ninety-three full-length videos with centrally-read eMS were obtained from 249 patients with ulcerative colitis, who participated in a phase II trial evaluating mirikizumab (NCT02589665). A video annotation approach was established to extract mucosal features and associated eMS classification labels from each video to be used as inputs for model training. The primary objective of the model was a categorical prediction of inactive vs active endoscopic disease evaluated against 2 independent test sets: a full set with a baseline single human expert read and a consensus subset in which 2 human reads agreed.ResultsOn the full test set of 147 videos, the model predicted inactive vs active endoscopic disease via the eMS with an area under the curve of 89%, accuracy of 84%, positive predictive value of 80%, and negative predictive value of 85%. In the consensus test set of 94 videos, the model predicted inactive vs active endoscopic disease with an area under the curve of 92%, accuracy of 89%, positive predictive value of 87%, and negative predictive value of 90%.ConclusionWe have demonstrated that this machine learning model supervised by mucosal features and eMS video annotations accurately differentiates key levels of endoscopic disease activity.
| S-EPMC11307476 | biostudies-literature
Project description:BackgroundUlcerative colitis (UC) is a chronic lifelong disease. The disease extent of UC can progress over time. This study aimed to assess whether cumulative inflammatory burden (CIB) is associated with disease extension in distal UC (proctitis [E1] and left-sided colitis [E2]) patients, and to develop a quantified indicator of CIB.MethodsIn this retrospective study based on a prospective registry, distal UC patients receiving colonoscopies in Xijing Hospital (Xi'an, China) from January 2000 to May 2019 were studied. We developed a new score, namely the time-adjusted average Mayo endoscopic score (TA-MES), calculated as dividing the sum of the cumulative average MES over a period of surveillance time by the length of the endoscopic examination interval, to quantify the CIB. Cox regression was used to identify other potential risk factors.ResultsA total of 295 UC patients were followed for 1,487.02 patient-years. Among them, 140 patients (47.5%) experienced disease extension. Multivariate analysis showed that the TA-MES was significantly associated with disease extension in E1 (hazard ratio [HR], 2.90; 95% confidence interval [CI], 1.58-5.33, P = 0.001) and E2 (HR, 1.89; 95% CI, 1.16-3.09, P = 0.011) patients. Other risk factors included hemoglobin of <90 g/L and appendiceal skip inflammation; the protective factors included age, E2 at diagnosis, former smoking, and 5-aminosalicylic acid dose. Otherwise, MES at diagnosis, maximal MES, and mean MES failed to estimate the risk of disease extension.ConclusionTA-MES is a good quantified indicator of CIB and is independently associated with increased disease extension in distal UC patients. Whether the dynamic multiple scoring system could be used as a risk factor in other chronic relapsing-remitting diseases is a direction for future research.
| S-EPMC9114755 | biostudies-literature
Project description:BackgroundMucosal healing (MH) evaluated by endoscopy is a novel target of therapy in UC as it is associated with improved long-term outcomes. It is defined based on the Mayo endoscopic score (MES), but it is still to define whether a value of MES 0 or 1 should be the target. The purpose of this paper is to present the results of a systematic review with meta-analysis which compares long-term outcomes of patients in steroid-free clinical remission with MES 0 with those with MES 1.MethodsA systematic electronic search of the literature was performed using Medline, Scopus, and CENTRAL through December 2020 (PROSPERO n:CRD42020179333). The studies concerned UC patients, in steroid-free clinical remission, with MES of 0 or 1, and with at least 12-months of follow-up.ResultsOut of 4611 citations, 15 eligible studies were identified. Increases in clinical relapse among patients with MES 1 were observed in all the studies included in this review, suggesting that MES of 1 have a higher risk of relapse than a score of 0. MES 0 patients displayed a lower risk of clinical relapse (OR 0.33; 95% CI 0.26-0.43; I2 13%) irrespective of the follow-up time (12-months or longer). On the other hand, no differences were found comparing MES 0 versus MES 1 about the risk of hospitalization or colectomy.ConclusionsMES 0 is associated with a lower rate of clinical relapse than is MES 1. For this reason, MES 0, rather than MES 0-1, should be considered the therapeutic target for patients with UC.
| S-EPMC8895505 | biostudies-literature
Project description:BackgroundData on the relative risk of malignant transformation in ulcerative colitis (UC) are insufficient. We investigated the potential value of the Mayo endoscopic score (MES) for predicting malignant transformation in patients with UC.MethodsData of patients with UC evaluated at our institute from June 1986 to December 2019 were retrospectively analyzed. The MES used in the study indicated the results of the first colonoscopy after hospitalization. We defined MES of 0-1 as low and MES of 2-3 as high. Univariable and multivariate logistic regression models were used for statistical analysis.ResultsAmong the 280 eligible patients with UC with a median follow-up time of 14 (interquartile range, 10.0-18.0) years, those with a high MES were more likely to develop malignant transformation. High MES positively correlated with the degree of malignancy and was an independent risk factor for UC-associated dysplasia and colorectal cancer (CRC, odds ratio [OR], 9.223; 95% confidence interval [CI], 1.160-73.323; p = 0.036). Disease duration >5 years (OR, 2.05; 95% CI, 1.177-3.572; p = 0.011), immunomodulator use (OR, 4.314; 95% CI, 1.725-10.785; p = 0.002), biologics nonuse (OR, 3.901; 95%CI, 2.213-6.876; p < 0.001), and Hb <90 g/L (OR, 2.691; 95% CI, 1.251-5.785; p = 0.011) were contributing factors for high MES.ConclusionHigh MES could be a novel predictor of malignant transformation in UC. Clinicians should optimize the use of biologics and immunomodulators early and should actively correct anemia to improve the MES and then reduce the incidence of UC-associated dysplasia and CRC.
| S-EPMC8965860 | biostudies-literature
Project description:ObjectivesThe subjectivity of the Physician Global Assessment (PGA) is a limitation of the Mayo score in assessing severity of ulcerative colitis (UC). We compared treatment efficacy using endpoint definitions based on modified Mayo (mMayo) score, versus those based on Mayo score, using data from the tofacitinib OCTAVE program.DesignThis post hoc analysis included data from two 8-week induction studies (OCTAVE Induction 1 and 2) and a 52-week maintenance study (OCTAVE Sustain).MethodsRemission and clinical response [with nonresponder imputation (NRI)] were assessed using mMayo (without PGA) and Mayo scores, and further stratified by prior tumor necrosis factor inhibitor (TNFi) failure status.ResultsAt week 8 of OCTAVE Induction 1 and 2, remission rates with placebo and tofacitinib 10 mg twice daily (BID), respectively, were 7.7% and 24.8% (mMayo) and 6.0% and 17.6% (Mayo). At week 52 of OCTAVE Sustain, remission rates with placebo, tofacitinib 5 and 10 mg BID, respectively, were 12.1%, 35.9%, and 42.1% (mMayo) and 11.1%, 34.3%, and 40.6% (Mayo). A statistically significant (p < 0.05) treatment effect of tofacitinib versus placebo was observed for remission and clinical response at all time points, regardless of scoring definition or prior TNFi failure status.ConclusionsA significant effect of tofacitinib versus placebo was demonstrated across efficacy endpoints using mMayo score, consistent with previously reported data using Mayo score. Treatment effect sizes were generally similar regardless of scoring definition. This observation may help contextualize tofacitinib therapy outcomes with those of new UC therapies and support the use of Mayo score-based endpoints in UC clinical trials.Trail registrationClinicalTrials.gov identifiers: NCT01465763; NCT01458951; NCT01458574.
| S-EPMC9726836 | biostudies-literature
Project description:BackgroundFecal calprotectin is widely used to monitor disease activity in patients with inflammatory bowel disease. Multiple commercial kits exist, however, since the analyses are not standardized, these kits cannot be used interchangeably. We aimed to perform a technical evaluation of two kits (Calpro from Calprolab, Norway and Calprest from Eurospital, Italy) and perform a tuning for detection of clinically relevant disease states in ulcerative colitis.Materials and methodsFor tuning against different clinical states a total of 116 patients with ulcerative colitis were recruited (67 of which were part of an earlier publication). For the technical evaluation an additional series of 80 random samples from the hospital lab were included. Technical evaluation was done by correlation and limits of agreement analysis; cut-off levels were explored by ROC analysis against clinically relevant actual states.ResultsThe technical evaluation revealed good correlation between assays, however a non-linear difference was found: At values below 200 mg/kg, no significant bias was found; in the interval 200-1000 mg/kg the Calprest assay measured on average 30% lower than Calpro; and at higher values Calprest measured 60% higher values than Calpro. Both assays predicted Mayo endoscopic score (MES) 0 (cutoff 28: sensitivity 0.38; specificity 0.82 for Calprest; cutoff 28: sensitivity 0.50; specificity 0.77 for Calpro), and MES 2-3 (cutoff 148: sensitivity 0.72; specificity 0.80 for Calprest; cutoff 208: sensitivity 0.64; specificity 0.80 for Calpro), but did not predict normalization of mucosal TNF transcript per se. A combination of calprotectin and MES predicted mucosal TNF transcript values reasonably well (Calpro: sensitivity 0.85, specificity 0.58; Calprest: sensitivity 0.85, specificity 0.61).ConclusionThe Calpro and Calprest assays correlated well, but subtle differences were found, underlining the need for kit-specific cut-off values. Both kits were most precise in predicting active inflammation (MES 2-3), but less so for prediction of mucosal healing (MES 0) and normalization of mucosal TNF gene expression.
| S-EPMC6886764 | biostudies-literature
Project description:The Mayo score and a noninvasive 9-point partial Mayo score are used as outcome measures for clinical trials assessing therapy for ulcerative colitis (UC). There are limited data assessing what defines a clinically relevant change in these indices. We sought to assess what constitutes a clinically meaningful change in these indices using data from a recently completed placebo-controlled clinical trial.In all, 105 patients were enrolled in a 12-week randomized, placebo-controlled trial assessing rosiglitazone for treatment of mild to moderate UC. We compared the change in the Mayo score, the partial Mayo score, and a 6-point score composed just of the stool frequency and bleeding components of the Mayo score to the patient's perception of disease activity at week 0 and week 12. Optimal cutpoints were calculated as the maximal product of sensitivity and specificity.Each index was strongly correlated with the patient's rating of disease activity at week 12 (Spearman correlations 0.61-0.71, P < 0.0001 for all correlations). The maximal product of sensitivity and specificity to identify patient reported improvement of disease activity was achieved using cutpoints for change of 2.5 for the Mayo score (sensitivity 88%, specificity 80%), 2.5 for the partial Mayo score (sensitivity 88%, specificity 87%), and 1.5 for the 6-point score (sensitivity 88%, specificity 80%).The partial Mayo score and the 6-point score composed solely of the stool frequency and bleeding components performed as well as the full Mayo score to identify patient perceived clinical response.
| S-EPMC2597552 | biostudies-literature
Project description:BackgroundTofacitinib is an oral small molecule Janus kinase inhibitor for the treatment of ulcerative colitis. We evaluated tofacitinib efficacy and safety in the 52-week maintenance study, OCTAVE Sustain, by baseline Mayo endoscopic subscore (MES) following 8-week induction.MethodsThe proportion of patients achieving efficacy endpoints at Week 24 or 52 of OCTAVE Sustain was evaluated by baseline MES following 8-week induction. Using logistic regression, the difference in treatment effect (tofacitinib vs. placebo) between baseline MES (0 vs. 1) for each endpoint was assessed. Adverse events were evaluated.ResultsAt Week 52 of OCTAVE Sustain, a numerically higher proportion of tofacitinib-treated patients achieved remission with OCTAVE Sustain baseline MES of 0 versus 1 (61.9% vs. 36.5% for tofacitinib 5 mg twice daily [BID] and 75.0% vs. 54.2% for tofacitinib 10 mg BID). Similar trends were observed for endoscopic remission and endoscopic improvement. Logistic regression analyses showed a larger treatment effect at Week 52 in patients with baseline MES of 0 versus 1 for clinical response (p = 0.0306) in the tofacitinib 5 mg BID group (other endpoints all p > 0.05); differences were not significant for any endpoint in the 10 mg BID group (all p > 0.05). Infection adverse events were less frequent among patients with baseline MES 0 versus 1.ConclusionsMES may be important in predicting long-term efficacy outcomes for tofacitinib maintenance treatment. Aiming for endoscopic remission during induction with tofacitinib 10 mg BID may allow successful maintenance with tofacitinib 5 mg BID. Safety was consistent with the known tofacitinib safety profile. Trial registration NCT01458574.
| S-EPMC9906955 | biostudies-literature
Project description:IntroductionThe Mayo Clinic Endoscopic Subscore is a commonly used grading system to assess the severity of ulcerative colitis. Correctly grading colonoscopies using the Mayo Clinic Endoscopic Subscore is a challenging task, with suboptimal rates of interrater and intrarater variability observed even among experienced and sufficiently trained experts. In recent years, several machine learning algorithms have been proposed in an effort to improve the standardization and reproducibility of Mayo Clinic Endoscopic Subscore grading.MethodsHere we propose an end-to-end fully automated system based on deep learning to predict a binary version of the Mayo Clinic Endoscopic Subscore directly from raw colonoscopy videos. Differently from previous studies, the proposed method mimics the assessment done in practice by a gastroenterologist, that is, traversing the whole colonoscopy video, identifying visually informative regions and computing an overall Mayo Clinic Endoscopic Subscore. The proposed deep learning-based system has been trained and deployed on raw colonoscopies using Mayo Clinic Endoscopic Subscore ground truth provided only at the colon section level, without manually selecting frames driving the severity scoring of ulcerative colitis.Results and conclusionOur evaluation on 1672 endoscopic videos obtained from a multisite data set obtained from the etrolizumab Phase II Eucalyptus and Phase III Hickory and Laurel clinical trials, show that our proposed methodology can grade endoscopic videos with a high degree of accuracy and robustness (Area Under the Receiver Operating Characteristic Curve?=?0.84 for Mayo Clinic Endoscopic Subscore???1, 0.85 for Mayo Clinic Endoscopic Subscore???2 and 0.85 for Mayo Clinic Endoscopic Subscore???3) and reduced amounts of manual annotation.Plain language summaryPatient, caregiver and provider thoughts on educational materials about prescribing and medication safety Artificial intelligence can be used to automatically assess full endoscopic videos and estimate the severity of ulcerative colitis. In this work, we present an artificial intelligence algorithm for the automatic grading of ulcerative colitis in full endoscopic videos. Our artificial intelligence models were trained and evaluated on a large and diverse set of colonoscopy videos obtained from concluded clinical trials. We demonstrate not only that artificial intelligence is able to accurately grade full endoscopic videos, but also that using diverse data sets obtained from multiple sites is critical to train robust AI models that could potentially be deployed on real-world data.
| S-EPMC7917417 | biostudies-literature
Project description:ImportanceAssessing endoscopic disease severity in ulcerative colitis (UC) is a key element in determining therapeutic response, but its use in clinical practice is limited by the requirement for experienced human reviewers.ObjectiveTo determine whether deep learning models can grade the endoscopic severity of UC as well as experienced human reviewers.Design, setting, and participantsIn this diagnostic study, retrospective grading of endoscopic images using the 4-level Mayo subscore was performed by 2 independent reviewers with score discrepancies adjudicated by a third reviewer. Using 16 514 images from 3082 patients with UC who underwent colonoscopy at a single tertiary care referral center in the United States between January 1, 2007, and December 31, 2017, a 159-layer convolutional neural network (CNN) was constructed as a deep learning model to train and categorize images into 2 clinically relevant groups: remission (Mayo subscore 0 or 1) and moderate to severe disease (Mayo subscore, 2 or 3). Ninety percent of the cohort was used to build the model and 10% was used to test it; the process was repeated 10 times. A set of 30 full-motion colonoscopy videos, unseen by the model, was then used for external validation to mimic real-world application.Main outcomes and measuresModel performance was assessed using area under the receiver operating curve (AUROC), sensitivity and specificity, positive predictive value (PPV), and negative predictive value (NPV). Kappa statistics (κ) were used to measure agreement of the CNN relative to adjudicated human reference cores.ResultsThe authors included 16 514 images from 3082 unique patients (median [IQR] age, 41.3 [26.1-61.8] years, 1678 [54.4%] female), with 3980 images (24.1%) classified as moderate-to-severe disease by the adjudicated reference score. The CNN was excellent for distinguishing endoscopic remission from moderate-to-severe disease with an AUROC of 0.966 (95% CI, 0.967-0.972); a PPV of 0.87 (95% CI, 0.85-0.88) with a sensitivity of 83.0% (95% CI, 80.8%-85.4%) and specificty of 96.0% (95% CI, 95.1%-97.1%); and NPV of 0.94 (95% CI, 0.93-0.95). Weighted κ agreement between the CNN and the adjudicated reference score was also good for identifying exact Mayo subscores (κ = 0.84; 95% CI, 0.83-0.86) and was similar to the agreement between experienced reviewers (κ = 0.86; 95% CI, 0.85-0.87). Applying the CNN to entire colonoscopy videos had similar accuracy for identifying moderate to severe disease (AUROC, 0.97; 95% CI, 0.963-0.969).Conclusions and relevanceThis study found that deep learning model performance was similar to experienced human reviewers in grading endoscopic severity of UC. Given its scalability, this approach could improve the use of colonoscopy for UC in both research and routine practice.
| S-EPMC6537821 | biostudies-literature