Computer-aided classification of colorectal segments during colonoscopy: a deep learning approach based on images of a magnetic endoscopic positioning device

OBJECTIVE

Assessment of the anatomical colorectal segment of polyps during colonoscopy is important for treatment and follow-up strategies, but is largely operator dependent. This feasibility study aimed to assess whether, using images of a magnetic endoscope imaging (MEI) positioning device, a deep learning approach can be useful to objectively divide the colorectum into anatomical segments.

METHODS

Models based on the VGG-16 based convolutional neural network architecture were developed to classify the colorectum into anatomical segments. These models were pre-trained on ImageNet data and further trained using prospectively collected data of the POLAR study in which endoscopists were using MEI (3930 still images and 90,151 video frames). Five-fold cross validation with multiple runs was used to evaluate the overall diagnostic accuracies of the models for colorectal segment classification (divided into a 5-class and 2-class colorectal segment division). The colorectal segment assignment by endoscopists was used as the reference standard.

RESULTS

For the 5-class colorectal segment division, the best performing model correctly classified the colorectal segment in 753 of the 1196 polyps, corresponding to an overall accuracy of 63%, sensitivity of 63%, specificity of 89% and kappa of 0.47. For the 2-class colorectal segment division, 1112 of the 1196 polyps were correctly classified, corresponding to an accuracy of 93%, sensitivity of 93%, specificity of 90% and kappa of 0.82.

CONCLUSION

The diagnostic performance of a deep learning approach for colorectal segment classification based on images of a MEI device is yet suboptimal (clinicaltrials.gov: NCT03822390).

Lower Risk of Recurrence with a Higher Induction Dose of Mesalazine and Longer Duration of Treatment in Ulcerative Colitis: Results from the Dutch, Non-Interventional, IMPACT Study

BACKGROUND AND AIMS

The dose and duration of mesalazine treatment for ulcerative colitis (UC) is a potentially important determinant of effectiveness, with evidence suggesting that continuing the induction dose for 6-12 months may improve outcomes; however, real-world data are lacking. We assessed mesalazine use in Dutch clinical practice, including how differences in dose and duration affected UC outcomes.

METHODS

Adults with mild-to-moderate UC who received oral prolonged-release mesalazine de novo or had a dose escalation for an active episode were followed for 12 months in this non-interventional study (ClinicalTrials.gov identifier: NCT02261636). The primary endpoint was time from start of treatment to dose reduction (TDR). Secondary endpoints included recurrence rate, adherence, and work productivity.

RESULTS

In total, 151 patients were enrolled, of whom 108 (71.5%) were newly diagnosed with UC. The majority (120; 79.5%) received a dose of ≥4 g/day. Nearly one-third (48; 31.8%) underwent dose reduction, with mean TDR being 8.3 months. Disease extent and endoscopic appearance did not influence duration of induction therapy, while TDR increased with higher baseline UCDAI scores. TDR was longer in patients without (mean 8.8 months) than with (4.1 months) recurrence, although not significantly (p=0.09). Patients on ≥4 g/day had a significantly lower chance of recurrence versus those on 2-<4 g/day (26.6% vs 62.5%, respectively; p=0.04). Longer treatment duration was associated with significantly reduced recurrence risk [hazard ratio >6 months vs 3-6 months: 0.19 (95%CI: 0.08-0.46); p<0.05], particularly for those on ≥4 g/day [0.15 (0.06-0.40) vs 0.26 (0.01-11.9) for 2-<4 g/day). Patients reported significantly increased work productivity, which was maintained throughout follow-up.

CONCLUSIONS

Mesalazine was effective induction therapy, with treatment duration not meaningfully influenced by disease extent and endoscopic appearance at initiation. A higher induction dose of oral mesalazine (≥4 g/day) and longer duration of treatment (>6 months) was associated with a lower recurrence risk.

Implications for a role of interleukin-23 in the pathogenesis of chronic gastritis and of peptic ulcer disease

The present study aimed to investigate the role of gastric mucosa for the secretion of interleukin (IL)-23 in chronic gastritis. One hundred and one patients were enrolled; 47 with duodenal ulcer, 33 with gastric ulcer and 31 with chronic gastritis. Biopsies were incubated in the absence/presence of endotoxins. Supernatants were collected and IL-23 and IL-1beta were measured by enzyme-linked immunosorbent assay. Scoring of gastritis was performed according to the updated Sydney score. Patients with duodenal and gastric ulcer and those with chronic gastritis had similar scores of gastritis. IL-23 was higher in supernatants of tissue samples of Helicobacter pylori-positive than of H. pylori-negative patients. No differences were recorded in concentrations of IL-23 and IL-1beta between patients with duodenal ulcer, gastric ulcer and chronic gastritis. Positive correlations were found between IL-23 of patients with both duodenal and gastric ulcer and chronic gastritis and the degree of infiltration of neutrophils and monocytes. Similar correlations were observed between IL-23 and IL-1beta. IL-23 secreted by the gastric mucosa could be implicated in the pathogenesis of chronic gastritis. IL-23 was released in the presence of H. pylori from the inflamed gastric mucosa and followed the kinetics of IL-1beta.