Artificial Intelligence-Assisted Endoscopy in Ulcerative Colitis

Artificial Intelligence-Assisted Endoscopy In Ulcerative Colitis

Inflammatory bowel diseases (IBD) are disorders that cause chronic inflammation in the gastrointestinal (GI) tract. The inflammation is the result from dysregulated immune response to environmental triggers in genetically susceptible hosts. The two most common forms of IBD are Crohn’s disease and Ulcerative Colitis (UC) (De Souza & Fiocchi, 2016).

Lower GI endoscopy (colonoscopy) is of paramount importance in the management of IBD. Imaged by high-definition video-camera via the colonoscope the mucosa of the colon is recorded and examined by the endoscopist. The colonoscopy is used for the diagnosis, the assessment of the disease severity, the evaluation of response to treatment and for dysplasia surveillance. Endoscopy is the gold standard method of discerning the disease severity and the treatment outcome in patients with UC. Determining the severity and the extent of the disease is important in guiding the management.

In ulcerative colitis, the endoscopic appearance of the mucosa is depending on the degree of inflammation. Therefore, based on the endoscopic features of the inflammation, endoscopic grading scores are used to assess the disease severity and monitor the response to therapy (Lewis et al., 2008; Walsh, Bryant, & Travis, 2016).

The total Mayo score is the most used tool to assess the severity of ulcerative colitis activity. The score is the sum of the individual scores of clinical symptoms including the rectal bleeding and stool frequency, the physician’s global assessment, and the endoscopic appearance of the colonic mucosa (Lewis et al., 2008; Walsh et al., 2016).

The Mayo endoscopic subscore (MES) describes the severity of mucosal inflammation by evaluating the mucosal vascular pattern status, the presence of mucosal erosions or ulcerations, and the presence and degree of mucosal friability or bleeding (Walsh et al., 2016). The mucosal appearance at endoscopy defines the MES score; 0 = normal or inactive disease, 1 = mild disease (erythema, decreased vascular pattern, mild friability, 2 = moderate disease (marked erythema, absent vascular pattern, friability, erosions), 3 = severe disease (spontaneous bleeding, ulceration).

Determining the severity and the extent of the disease is important in guiding the management. Moreover, the clinical and endoscopic remission, where there is complete resolution of the endoscopic lesions (defined as mucosal healing), is currently the recommended treatment target in ulcerative colitis since it is associated with improved long-term outcomes (Peyrin-Biroulet et al., 2015). Recent studies have shown that complete mucosal healing predicts sustained clinical remission, the need for hospitalization and the probability of surgery-free survival among patients (Colombel et al., 2011). Therefore, a precise and detailed real-time assessment of the mucosal surface has become very important for the medical management of IBD patients.

On the other hand, the interpretation of the endoscopic findings is operator dependent. Moreover, in clinical practice, the use of the endoscopic scores is compromised by the facts that they are numerous and complex and that the majority of non-IBD dedicated gastroenterologists are not familiar with them. The use of the MES is subject of inter-individual differences resulting in substantial inter-observer and intra-observer variability in the grading of the endoscopic severity, even among experts, which can affect the therapeutic management of the IBD patients (Daperno et al., 2014; Osada et al., 2010).

The development of training programs for the inexperienced gastroenterologists could improve and maintain their proficiency in using the IBD endoscopic scoring systems (Daperno et al., 2017). Central reading by trained experts is a strategy which is used in clinical trials, but it is time consuming, expensive and it is not practical for the everyday patient care.

To address this problem, one solution is the use of a digital endoscopic image or video analysis system, which enables computer to understand endoscopic data for an intelligent and automated disease diagnosis and severity classification.