Page 105 - Read Online

Page 8 of 11 Jalal et al. J Cancer Metastasis Treat 2023;9:11 https://dx.doi.org/10.20517/2394-4722.2022.122

Attempts have been made to incorporate artificial intelligence (AI) technology into clinical practice to aid
the management of PCLs and reduce human bias. Kurita et al. obtained a high diagnostic accuracy (92.9%)
by constructing a diagnostic algorithm using relevant variables: CEA, CA 125, a cystic fluid amylase, the
[67]
type of cyst, sex, cyst location, the connection of the pancreatic duct to the cyst, and the type of cyst . This
type of AI algorithm relies on the quality and availability of the relevant information that is fed into it. Its
functionality may be impaired if information is missing. Kuwahara et al. obtained a diagnostic accuracy of
86.2% in detecting malignancy in IPMNs via deep learning using EUS images . In addition, their study
[68]
yielded higher accuracy than human diagnosis or the presence of mural nodules; however, there was more
cancer compared to routine clinical practice. The study was limited by its sample size and retrospective
nature. By using clinical, imaging, and molecular markers from 436 patients, another retrospective study
developed a supervised learning software programme to categorise PCLs into those that require surgery or
surveillance and those that can be safely discharged . Using histopathology as a gold standard, the clinical
[69]
management outcome from using the software was more accurate than from standard care.

EUS-guided pancreatic cyst ablation using ethanol and/or paclitaxel has been investigated for the non-
surgical treatment of PCLs. It can provide a minimally invasive treatment, especially for those deemed at
high risk for surgical resection. A small study (n = 13) reported complete or partial resolution in 12
[70]
patients . A larger study (n = 162) confirmed its high efficacy in unilocular and small cysts with complete
or partial resolution in 91.8%. During the follow-up period, out of 114 patients (post-ablation), only two had
cyst recurrences in six years .
[71]

PCLs have stimulated considerable research on developing new imaging, endoscopic, and sampling
techniques. The approach to each patient should be individualised based on their clinical status, the
presence of co-morbidities, and the risk of malignancy. MRI may help to identify mucinous cysts or
malignant changes. However, a combination of imaging morphology with EUS-FNA may be required.
Cystic fluid biomarkers can be used as an adjunct in predicting high-risk PCLs and those requiring early
surgical resections. TTNB, nCLE, and pancreatoscopy are promising new tests that can be used in the
assessment and management of PCLs. Pancreatoscopy should be reserved for equivocal cases due to the
high risk of pancreatitis. Some of the advances reported in this review may not be applicable in clinical
practice due to their limited availability. Therefore, the classic high-risk physical symptoms remain the most
important and valuable predictors of surgical therapy.

Future research is required to develop non-invasive tests and markers that can differentiate benign from
premalignant or malignant PCLs.

DECLARATIONS
Authors’ contributions
Made substantial contributions to the conception and design of the article and interpreting the relevant
literature: Jalal M, Gbadegesin S, Ibrahim S, Tehami N
Drafted the article or revised it critically for important intellectual content: Jalal M, Gbadegesin S, Ibrahim
S, Tehami N

Availability of data and materials
Not applicable.

Financial support and sponsorship
None.