A Sessile Serrated Lesion Overlying a Submucosal Colonic Lipoma: An Endoscopic Rarity Identified Using Artificial Intelligence

Background

First described by Bauer in 1757, lipomas are slow-growing mesenchymal neoplasms, uncommon in the colon, and most often identified as an incidentaloma tumor in asymptomatic patients during colonoscopy [1,2]. Because its low clinical relevance, they receive little attention in the literature.

Lipomas are generally found in the right colon [3], and the ascending colon accounts for 40–85% [4]. Although rare, when larger than 3 cm in size they may present symptoms such as abdominal pain, change in bowel habits, bleeding, or even intestinal intussusception, requiring emergency surgery [5].

They are more prevalent in females, with a peak incidence in the fifth to sixth decades of life, rarely overlapping with epithelial lesions [1].

Mostly, lipomas do not need to be resected but great attention should be paid to the overlying mucosa. Most reports refer to isolated findings of association between lipoma and adenomatous polyp [6–8]. Although very rare, cases of lipomas in association with hyperplastic polyps have also been described [9].

Atypical lipomas were first described by Snover in 1984 [10], referring to lipomas with malignant characteristics such as cytological changes in adipose cells, with a sarcomatous appearance.

Artificial intelligence has been used as an additional tool for diagnosis and for characterization of colonic lesions. It has been demonstrated that AI increases the adenoma detection rate [11].

Our aim here was to describe a rare case of a sessile serrated lesion (SSL) overlying a submucosal lipoma identified with the aid of artificial intelligence in a patient undergoing colonoscopy screening.

Case Report

A 60-year-old woman who was an employee of the University Medical School, completely asymptomatic, in a colorectal cancer screening program, during her periodic medical examination, was submitted to a fecal immunochemical test (FIT), with a result of 64 ng/mL (reference <50 ng/mL). Since her FIT was positive, she was referred for colonoscopy. She had high blood pressure, dyslipidemia, and regular use of medication, and she denied smoking and family history of colorectal cancer or previous abdominal surgeries.

A high-definition colonoscope (Fujifilm Eluxeo 7000/760) assisted by artificial intelligence (AI) (Fujifilm CAD EYE) was used. The entire colon was in good preparation condition, considered as grade 9 on the Boston Bowel Preparation Scale.

At the ascending colon, an elevated flat lesion, Paris 0-IIa type, measuring approximately 7 mm in size, was simultaneously observed by white-light endoscopy and AI. AI characterized it as hyperplastic. A type II-O crypt opening pattern described by Kimura was observed, suggestive of a sessile serrated lesion (Figure 1).

After injection of 0.9% saline solution at the submucosa, the lesion was resected en bloc using a diathermic snare, at which time exposure of adipose tissue was noted. At first, a possible perforation was considered, but after careful review of the base of the wound, this hypothesis was ruled out.

Upon finding submucosal fat after polyp excision and discarding perforation, the complete lipoma was not re-excised. The lipoma appeared to be twice as large as the lesion.

Histopathological examination of the specimen revealed a submucous lipoma associated with a sessile serrated lesion. The epithelium was serrated and composed of columnar and goblet cells, but without atypia or mitotic activity (Figure 2).

Three other colonic polyps were also diagnosed with the aid of AI, all of them characterized as neoplastic: An 8-mm sessile polyp in the transverse colon; a 5-mm sessile descending polyp; and a 15-mm pedunculated polyp in the sigmoid colon (Figure 3). Subsequently, pathological anatomy confirmed these findings, all of them diagnosed as tubular adenomas with low-grade dysplasia.

Clinical follow-up colonoscopy after 1 year demonstrated scar tissue with no residual lesion.

Discussion

We present this case to call attention to use of AI in diagnosis of flat lesion. Since the association of a serrated lesion with a lipoma is rare and might indicate an atypical lipoma, the correct diagnosis is imperative for adequate management.

Atypical lipomas are lipomas with malignant characteristics related to cytological changes in adipose cells with a sarcomatous appearance [10]. However, in 2016, Virgilio et al [12] proposed a new classification of atypical lipomas, in which the liposarcomatous changes described by Snover were found, along with a potential association with adenocarcinoma.

It is not clear if the association of a serrated polyp with underlying lipoma is a coincidence or if there is a pathogenic correlation between these entities [13].

According to Virgilio et al [12], the finding of precancerous lesions on a lipoma, considered as an atypical lipoma, is a malignant transformation in the overlying mucosal epithelium and should be completely resected.

Therefore, lipomas can be classified based on the histological appearance, which may indicate some relationship with their clinicopathological presentation and may suggest the presence of a genetic syndrome [14].

As in our case, in which we initially thought of a perforation immediately after resection due to a “naked fat” sign, similar suspected cases have been reported [15,16]. However, after closer evaluation of the wound, an intact muscle layer was observed.

It is difficult to determine the actual size of the lipoma, as the lipoma is deep in the submucosa. The submucosal lipoma was not suggested by the AI and we believe that this was probably due to the flat appearance of the lipoma.

Nowadays, we know through epigenetics that the serrated pathway is responsible for one-third of all colorectal cancers [17].

Capra et al [18] reported an adenoma on a lipoma in the left colon, suggesting that constant trauma caused by stools on the colonic mucosa could be the cause of this potential malignancy. Such a hypothesis is be supported in our case, as the lesion was in the right colon, where there are no fully formed stools.

Several previous studies have already demonstrated that the detection of serrated polyps varies widely among endoscopists, which leads to underdiagnosis of these lesions [19,20].

Despite characterization of the lesions by AI into hyper-plastic and neoplastic, AI so far cannot distinguish between hyperplastic and serrated polyps, but it helps identify small polyps that may not be diagnosed by less experienced clinicians. In this case report, the colonoscopy was performed by an experienced physician, and no polyps were missed by the physician and detected by AI.

Careful endoscopic examination together with high-tech endoscopy are imperative during colonoscopy, especially during colorectal cancer screening, to achieve better management and prognosis.

Conclusions

AI can improve diagnosis of complex or overlapping conditions, helping less experienced physicians or trainees. Use of AI in endoscopic images could help reduce overall costs by avoiding misdiagnosis or delayed treatment.